Skip to content
BranchesTags

Latest commit

 

History

44 Commits

Folders and files

NameName
Last commit message
Last commit date
docs/assets
docs/assets
 
 
LICENSE
LICENSE
 
 
README.md
README.md
 
 
index.html
index.html
 
 

Repository files navigation

Awesome Logo arXiv Project Page Visitors PR's Welcome

😎 Awesome AI Auto-Research

This repository accompanies the survey paper "AI for Auto-Research: Roadmap & User Guide" and tracks papers on AI-assisted and automated scientific research, covering the full research lifecycle.

🤖 AI Auto-Research

We organize the academic research lifecycle as eight interconnected stages grouped into four epistemological phases. Each phase serves a distinct function in producing, scrutinizing, and communicating scientific knowledge.
Phase 1: Creation
Generating novel research ideas, searching and synthesizing literature, running coding experiments, and creating publication-quality tables and figures. This phase spans Idea Generation, Literature Review, Coding & Experiments, and Tables & Figures.
Phase 2: Writing
Drafting, editing, and polishing academic manuscripts. AI assistance ranges from semi-automated grammar and citation tools to fully automated paper generation — the most commercially mature yet ethically contested stage.
Phase 3: Validation
Automated peer review generation, reviewer-paper matching, review quality assessment, and AI-assisted author rebuttals. This phase covers Peer Review and Rebuttal & Revision.
Phase 4: Dissemination
Converting papers into slides, posters, videos, websites, and social media content. Each output format targets a different audience and demands its own design logic and AI tool chain.

For additional details, kindly refer to our 📚 Paper and 🌏 Project Page.

📚 Citation

If you find this work helpful for your research, please kindly consider citing our paper:

@article{survey-ai-auto-research,
  title   = {{AI} for {Auto-Research}: Roadmap \& User Guide},
  author  = {Kong, Lingdong and Sun, Xian and Chow, Wei and Li, Linfeng and Lin, Kevin Qinghong and Zhang, Xuan Billy
             and Wang, Song and Li, Rong and Wu, Qing and Gao, Wei and Wang, Yingshuo and Xie, Shaoyuan
             and Liu, Jiachen and Qu, Leigang and Li, Shijie and Ng, Lai Xing and Cottereau, Benoit R.
             and Liu, Ziwei and Chua, Tat-Seng and Ooi, Wei Tsang},
  journal = {arXiv preprint arXiv:2605.18661},
  year    = {2026}
}

Table of Contents

1. Idea Generation

LLM Internal Knowledge-Based Generation

In chronological order, from the earliest to the latest.

Model Paper Venue Website GitHub
Chain of Ideas arXiv
Chain of Ideas: Revolutionizing Research Via Novel Idea Development with LLM Agents		 arXiv '24 - GitHub
ResearchAgent Website
ResearchAgent: Iterative Research Idea Generation over Scientific Literature with Large Language Models		 NAACL '25 - GitHub
SciMON arXiv
SciMON: Scientific Inspiration Machines Optimized for Novelty		 ACL '24 - GitHub
Idea Gen Agent arXiv
Can LLMs Generate Novel Research Ideas? A Large Scale Human Study with 100+ NLP Researchers		 arXiv '24 - -
IRIS Website
IRIS: Interactive Research Ideation System for Accelerating Scientific Discovery		 ACL '25 - GitHub
Spark arXiv
Spark: A System for Scientifically Creative Idea Generation		 ICCC '25 - -
Diverse Hypo. Search arXiv
Towards Diverse Scientific Hypothesis Search with Large Language Models		 arXiv '26 - -
Tree-of-Ideas arXiv
Tree-of-Ideas: Automated Research Ideation via Cross-Trajectory Reasoning over Scholarly Evolution		 arXiv '26 - -
IDEAgent arXiv
IDEAgent: Agentic Quality-Diversity Search for Research Idea Generation		 arXiv '26 - -

External Signal-Driven Generation

In chronological order, from the earliest to the latest.

Model Paper Venue Website GitHub
SGHA arXiv
SGHA: Evidence-Grounded Research Problem Discovery with Local Language Models		 arXiv '26 - -
MAIL arXiv
MAIL: Memory-driven, Adaptive, Incremental, and Literature-grounded Framework for Hypothesis Generation in Chemistry		 arXiv '26 - -
MOOSE-Chem Website
MOOSE-Chem: Large Language Models for Rediscovering Unseen Chemistry Scientific Hypotheses		 ICLR '25 - -
Nova arXiv
Nova: An Iterative Planning and Search Approach to Enhance Novelty and Diversity of LLM Generated Ideas		 arXiv '24 - -
SciAgents arXiv
SciAgents: Automating Scientific Discovery through Multi-Agent Intelligent Graph Reasoning		 arXiv '24 - GitHub
SciPIP arXiv
SciPIP: An LLM-based Scientific Paper Idea Proposer		 arXiv '24 - GitHub
IdeaSynth arXiv
IdeaSynth: Iterative Research Idea Development Through Evolving and Composing Idea Facets with Literature-Grounded Feedback		 CHI '25 - -
MOOSE-Chem2 Website
MOOSE-Chem2: Exploring LLM Limits in Fine-Grained Scientific Hypothesis Discovery via Hierarchical Search		 NeurIPS '25 - -
HALO arXiv
HALO: Interactive Co-abductive Reasoning in Scientific Hypothesis Generation		 arXiv '26 - -
TCA-SIR arXiv
TCA-SIR: Learning Target-Conditioned Abstractions for Scientific Inspiration Retrieval		 arXiv '26 - -
ECLAIR arXiv
ECLAIR: A Causally-Grounded AI Framework for Scientific Discovery in Empirical Software Engineering		 arXiv '26 - -

Multi-Agent Collaborative Generation

In chronological order, from the earliest to the latest.

Model Paper Venue Website GitHub
HypoForge arXiv
HypoForge: A Self-Improving Multi-Agent Framework for Automated Hypothesis Generation and Testing via Scientific Skill Learning		 arXiv '26 - -
Combi. Creativity arXiv
Combi. Creativity		 arXiv '24 - -
Deep Ideation arXiv
Deep Ideation: Designing LLM Agents to Generate Novel Research Ideas on Scientific Concept Network		 arXiv '25 - GitHub
VirSci Website
Many Heads Are Better Than One: Improved Scientific Idea Generation by A LLM-Based Multi-Agent System		 ACL '25 - GitHub
Multi-Agent Dial. arXiv
Multi-Agent Dial.		 SIGDIAL '25 - -
Artificial Hivemind arXiv
Artificial Hivemind: The Open-Ended Homogeneity of Language Models (and Beyond)		 NeurIPS '25 - -
Auditable AI Sci. arXiv
Toward Auditable AI Scientists: A Hypothesis Evolution Protocol for LLM Agents		 arXiv '26 - -
Diverse Personalized Ideation arXiv
Diversifying Personalized Research Ideation against AI-Induced Homogenization		 arXiv '26 - -

Novelty and Feasibility Assessment

In chronological order, from the earliest to the latest.

Model Paper Venue Website GitHub
RATIO arXiv
RATIO: A Benchmark for Retrieval Across Typed Ideation Operations in Scientific Literature		 arXiv '26 - -
Lit2Test arXiv
What Proves You Wrong: Benchmarking Language Models on Falsifiable Research Ideation		 arXiv '26 - -
Energy Scoring arXiv
Do LLMs Know a Good Hypothesis When They See One? Logit-Based Energy Scoring Outperforms Prompted LLM-as-Judge for Scientific Hypothesis Ranking		 arXiv '26 - -
Think-Probe-Respond arXiv
Think-Probe-Respond: Improving Large Language Models as Judges of Research Idea Novelty		 EMNLP '26 - -
IdeaBench Website
LiveIdeaBench: Evaluating LLMs' Scientific Creativity and Idea Generation with Minimal Context		 KDD '25 - -
LiveIdeaBench arXiv
LiveIdeaBench: Evaluating LLMs' Scientific Creativity and Idea Generation with Minimal Context		 arXiv '24 - -
AI Idea Bench 2025 arXiv
AI Idea Bench 2025: AI Research Idea Generation Benchmark		 arXiv '25 - GitHub
HeurekaBench arXiv
HeurekaBench: A Benchmarking Framework for AI Co-scientist		 ICLR '26 - GitHub
ResearchBench arXiv
ResearchBench: Benchmarking LLMs in Scientific Discovery via Inspiration-Based Task Decomposition		 ACL '26 - -
HindSight arXiv
HindSight: Evaluating LLM-Generated Research Ideas via Future Impact		 arXiv '26 - -
Rubric Rewards arXiv
Training AI Co-Scientists Using Rubric Rewards		 arXiv '25 - -
DeepInnovator arXiv
DeepInnovator: Triggering the Innovative Capabilities of LLMs		 arXiv '26 - GitHub
FlowPIE arXiv
FlowPIE: Test-Time Scientific Idea Evolution with Flow-Guided Literature Exploration		 arXiv '26 - -
SoundnessBench arXiv
SoundnessBench: Can Your AI Scientist Really Tell Good Research Ideas from Bad Ones?		 arXiv '26 - -
LLM-Judge Novelty arXiv
On the Limits of LLM-as-Judge for Scientific Novelty Assessment		 arXiv '26 - -
LigBench arXiv
LigBench: A Unified and Human-Aligned Benchmark for LLM-based Research Idea Generation		 arXiv '26 - -
Reconstruction arXiv
Reconstruction: A Blind Benchmark for Recovering Research Ideas from Pre-Publication Bibliographies		 arXiv '26 - -

2. Literature Review & Paper Search

Literature Retrieval

In chronological order, from the earliest to the latest.

Model Paper Venue Website GitHub
ITER arXiv
ITER: Interaction-Aware Retrieval for Agentic Search		 arXiv '26 - -
Multi-Aspect Retrieval arXiv
Can Retrievers Find the Same Paper from Different Aspects? A Multi-Aspect Full-Paper Scientific Retrieval Benchmark		 arXiv '26 - -
CiteME arXiv
CiteME: Can Language Models Accurately Cite Scientific Claims?		 arXiv '24 - -
LitLLM arXiv
LitLLM: A Toolkit for Literature Review with Large Language Models		 arXiv '24 - -
LitSearch arXiv
LitSearch: A Retrieval Benchmark for Scientific Literature Search		 arXiv '24 - GitHub
PaperQA2 arXiv
Language Agents Achieve Superhuman Synthesis of Scientific Knowledge		 arXiv '24 - GitHub
OpenResearcher arXiv
OpenResearcher: Unleashing AI for Accelerated Scientific Research		 EMNLP '24 - -
PaSa arXiv
PaSa: An LLM Agent for Comprehensive Academic Paper Search		 arXiv '25 - GitHub
Self-Evolving Retrieval arXiv
Towards Self-Evolving Agentic Literature Retrieval		 arXiv '26 - -
MasterSet arXiv
MasterSet: A Large-Scale Benchmark for Must-Cite Citation Recommendation in the AI/ML Literature		 arXiv '26 - -
Search, Inspect, Fetch arXiv
Search, Inspect, Fetch: Exploiting Structure-Aware Boolean Retrieval for Deep-Research Agents		 arXiv '26 - -
Rubric Reranker arXiv
Training Documents Reranker with Search Rubrics for Deep Research Agent		 arXiv '26 - -
Personalized DR Refinement arXiv
Personalized Deep Research Query Refinement with Graph-Scaffolded Evidence Grounding		 arXiv '26 - -

Survey & Related Work Generation

In chronological order, from the earliest to the latest.

Model Paper Venue Website GitHub
DAS arXiv
Deep Academic Survey: Stateful Agentic Closed-Loop Paradigm for Academic Survey Automation		 arXiv '26 - -
Tree-of-Concerns arXiv
Tree-of-Concerns: Hierarchical Multi-Agent Debate for Unstated-Limitation Extraction in Scientific Critique		 EMNLP '26 - -
ChatPaper Website
ChatPaper: Use LLM to summarize papers		 GitHub '23 - GitHub
PaperQA arXiv
PaperQA: Retrieval-Augmented Generative Agent for Scientific Research		 arXiv '23 - GitHub
AutoSurvey arXiv
AutoSurvey: Large Language Models Can Automatically Write Surveys		 arXiv '24 - GitHub
GPT Researcher Website
GPT Researcher: Autonomous Agent for Comprehensive Online Research		 GitHub '24 - GitHub
LLMs for Lit. Review arXiv
LLMs for Lit. Review		 arXiv '24 - -
STORM arXiv
Assisting in Writing Wikipedia-like Articles From Scratch with Large Language Models		 arXiv '24 - GitHub
Agentic AutoSurvey arXiv
Agentic AutoSurvey: Let LLMs Survey LLMs		 arXiv '25 - -
Citegeist arXiv
Citegeist: Automated Generation of Related Work Analysis on the arXiv Corpus		 arXiv '25 - -
IterSurvey arXiv
IterSurvey: Deep Literature Survey Automation with an Iterative Workflow		 arXiv '25 - GitHub
LiRA arXiv
LiRA: A Multi-Agent Framework for Reliable and Readable Literature Review Generation		 arXiv '25 - -
SurveyForge arXiv
SurveyForge: On the Outline Heuristics, Memory-Driven Generation, and Multi-dimensional Evaluation for Automated Survey Writing		 arXiv '25 - GitHub
SurveyG arXiv
SurveyG: A Multi-Agent LLM Framework with Hierarchical Citation Graph for Automated Survey Generation		 arXiv '25 - -
SurveyX arXiv
SurveyX: Academic Survey Automation via Large Language Models		 arXiv '25 - -
InteractiveSurvey arXiv
InteractiveSurvey: An LLM-based Personalized and Interactive Survey Paper Generation System		 arXiv '25 - GitHub
CiteLLM arXiv
CiteLLM: An Agentic Platform for Trustworthy Scientific Reference Discovery		 arXiv '26 - -
DeepSurvey arXiv
DeepSurvey: Enhancing Analytical Depth and Citation Reliability in Automated Survey Generation		 arXiv '26 - -
STRUCTSURVEY arXiv
STRUCTSURVEY: Structured Agentic Retrieval for Automated Survey Paper Generation		 arXiv '26 - -

Deep Research Agents

In chronological order, from the earliest to the latest.

Model Paper Venue Website GitHub
Crase arXiv
Structurally-bounded Agentic Graph Exploration for Evidence-Grounded Scholarly DeepSearch		 arXiv '26 - -
DeepWeaver arXiv
DeepWeaver: Bridging the Evidence Synthesis Gap in Open-Ended Question Answering		 arXiv '26 - -
AgentR arXiv
AgentR A Stateful and Recovery-Aware Software Architecture for LLM-based Auditable Workflows		 arXiv '26 - -
ASReview Website
An Open Source Machine Learning Framework for Efficient and Transparent Systematic Reviews		 Nature MI '21 - GitHub
CHIME arXiv
CHIME: LLM-Assisted Hierarchical Organization of Scientific Studies for Literature Review Support		 arXiv '24 - -
DeepResearch-Agent Website
DeepResearchAgent: A Hierarchical Multi-Agent System for Deep Research		 GitHub '25 - GitHub
DeerFlow Website
DeerFlow: A Deep Research Framework Orchestrating Sub-Agents, Memory, and Sandboxes		 GitHub '25 - GitHub
OpenScholar Website
OpenScholar: Synthesizing Scientific Literature with Retrieval-Augmented LMs		 Nature '26 - -
AutoAgent arXiv
AutoAgent		 arXiv '25 - -
Tongyi DeepResearch Website
Tongyi DeepResearch		 GitHub '25 - GitHub
O-Researcher arXiv
O-Researcher: An Open Ended Deep Research Model via Multi-Agent Distillation and Agentic RL		 arXiv '26 - -
OpenResearcher arXiv
OpenResearcher: Unleashing AI for Accelerated Scientific Research		 arXiv '26 - GitHub
AREX arXiv
AREX: Towards a Recursively Self-Improving Agent for Deep Research		 arXiv '26 - -
Predictive Navigation arXiv
Deep Research Pretraining via Predictive Navigation		 arXiv '26 - -
On-Device DR (4B) arXiv
On-Device Deep Research at 4B: Exposure Bounds Faithfulness, Retrieval Bounds Coverage		 arXiv '26 - -
Carnot arXiv
Carnot: Interpretable, Interactive, and Optimized Execution of Deep Research Queries		 VLDB '26 - -
Marginal Value Est. arXiv
Not Worth Another Token: Marginal Value Estimation for Efficient Deep Research Agents		 arXiv '26 - -
Retrieval-Aware Control arXiv
When Deep Research Agents Stagnate: Enhancing Reasoning with Retrieval-Aware Agent Control		 arXiv '26 - -
Analogical Deep Research arXiv
Analogical Deep Research: Retrieving and Integrating Historical Analogies for Foresight Analysis		 arXiv '26 - -
Plato-Bio arXiv
Plato-Bio: verification-first biological novelty screening with temporal rediscovery and structural benchmarks		 arXiv '26 - -
Albilich arXiv
Albilich: Steerable Proof-State Orchestration for LLM-Based Mathematical Research with CAS Integration		 arXiv '26 - -

Retrieval and Synthesis Quality Assessment

In chronological order, from the earliest to the latest.

Model Paper Venue Website GitHub
IBIS arXiv
From Inertia to Objectivity: Improving Deep Research Agents with Noise Isolation		 EMNLP '26 - -
Agent to Blame arXiv
Who is the Agent to Blame? Localizing Faithfulness and Citation Mistakes in Agentic Deep Research		 EMNLP '26 - -
DeepScholar-Bench arXiv
DeepScholar-Bench: A Live Benchmark and Automated Evaluation for Generative Research Synthesis		 arXiv '25 - GitHub
ReportBench arXiv
ReportBench: Evaluating Deep Research Agents via Academic Survey Tasks		 arXiv '25 - GitHub
IDRBench arXiv
IDRBench: Interactive Deep Research Benchmark		 arXiv '26 - -
ScholarGym arXiv
ScholarGym: Benchmarking Large Language Model Capabilities in the Information-Gathering Stage of Deep Research		 arXiv '26 - -
SciNetBench arXiv
SciNetBench: A Relation-Aware Benchmark for Scientific Literature Retrieval Agents		 arXiv '26 - -
AutoResearchBench arXiv
AutoResearchBench: Benchmarking AI Agents on Complex Scientific Literature Discovery		 arXiv '26 - -
PaperMind arXiv
PaperMind: Benchmarking Agentic Reasoning and Critique over Scientific Papers in Multimodal LLMs		 arXiv '26 - -
DRNOISE arXiv
DRNOISE: Benchmarking Deep Research Agents in Misleading Evidence Environments		 arXiv '26 - -
HiEviDR-Bench arXiv
HiEviDR-Bench: A Benchmark for Hierarchical Evidence Aggregation in Deep Research		 arXiv '26 - -
SciExplore arXiv
SciExplore: Evaluating Autonomous Agents from Scientific Navigation to Information Integration		 arXiv '26 - -
WANDR arXiv
WANDR: A Benchmark for Wide and Deep Research		 arXiv '26 - -
QA-to-DR Bench arXiv
From Simple QA to Deep Research: A Verifiable Benchmark Constructed through Iterative Task Evolution		 arXiv '26 - -

3. Coding & Experimentation

Code Generation

In chronological order, from the earliest to the latest.

Model Paper Venue Website GitHub
SWE-bench arXiv
SWE-bench: Can Language Models Resolve Real-World GitHub Issues?		 ICLR '24 - GitHub
SWE-agent arXiv
SWE-agent: Agent-Computer Interfaces Enable Automated Software Engineering		 arXiv '24 - GitHub
OpenHands arXiv
OpenHands: An Open Platform for AI Software Developers as Generalist Agents		 ICLR '25 - GitHub
SWE-bench Pro arXiv
SWE-Bench Pro: Can AI Agents Solve Long-Horizon Software Engineering Tasks?		 arXiv '25 - -
SWE-EVO arXiv
SWE-EVO: Benchmarking Coding Agents in Long-Horizon Software Evolution Scenarios		 arXiv '25 - -

Paper-to-Code

In chronological order, from the earliest to the latest.

Model Paper Venue Website GitHub
ReproAgent arXiv
ReproAgent: Contract-Guided Paper-to-Code Reproduction		 EMNLP '26 - -
DeepRepro arXiv
DeepRepro: State-Aware Subplanning for Paper-to-Code Reproduction in Evolving Repositories		 CIKM '26 - -
FunSearch Website
Mathematical Discoveries from Program Search with Large Language Models		 Nature '24 - GitHub
SciCode arXiv
SciCode: A Research Coding Benchmark Curated by Scientists		 arXiv '24 - GitHub
PaperBench arXiv
PaperBench: Evaluating AI's Ability to Replicate AI Research		 arXiv '25 - GitHub
PaperCoder arXiv
Paper2Code: Automating Code Generation from Scientific Papers in Machine Learning		 arXiv '25 - GitHub
ResearchCodeBench arXiv
ResearchCodeBench: Benchmarking LLMs on Implementing Novel ML Research Code		 arXiv '25 - -
SciReplicate-Bench arXiv
SciReplicate-Bench: Benchmarking LLMs in Agent-driven Algorithmic Reproduction from Research Papers		 arXiv '25 - GitHub

Experiment Execution & Orchestration

In chronological order, from the earliest to the latest.

Model Paper Venue Website GitHub
Praxist arXiv
Praxist: From Experimental Artifacts to Solution Lineages		 arXiv '26 - -
Skill-Based Baselines arXiv
Can Coding Agents Build Robust Baselines? A Skill-Based Approach for Automating the Medical Imaging Model-Development Pipeline		 MICCAI '26 - -
BioPlanner arXiv
BioPlanner: Automatic Evaluation of LLMs on Protocol Planning		 arXiv '23 - GitHub
CRISPR-GPT arXiv
CRISPR-GPT for Agentic Automation of Gene-Editing Experiments		 arXiv '24 - -
DS-Agent arXiv
DS-Agent: Automated Data Science by Empowering Large Language Models with Case-Based Reasoning		 arXiv '24 - GitHub
MLE-Bench arXiv
MLE-Bench: Evaluating Machine Learning Agents on Machine Learning Engineering		 arXiv '24 - -
MLAgentBench arXiv
MLAgentBench: Evaluating Language Agents on Machine Learning Experimentation		 arXiv '24 - GitHub
MLR-Copilot arXiv
MLR-Copilot: Autonomous Machine Learning Research based on Large Language Models Agents		 arXiv '24 - -
AIDE arXiv
AIDE: AI-Driven Exploration in the Space of Code		 arXiv '25 - -
AlphaEvolve arXiv
AlphaEvolve: A Coding Agent for Scientific and Algorithmic Discovery		 arXiv '25 - -
AutoReproduce arXiv
AutoReproduce: Automatic AI Experiment Reproduction with Paper Lineage		 arXiv '25 - GitHub
CURIE arXiv
Curie: Toward Rigorous and Automated Scientific Experimentation with AI Agents		 arXiv '25 - GitHub
MLGym arXiv
MLGym: A New Framework and Benchmark for Advancing AI Research Agents		 arXiv '25 - -
MLR-Bench arXiv
MLR-Bench: Evaluating AI Agents on Open-Ended Machine Learning Research		 arXiv '25 - -
Execution-Grounded arXiv
Towards Execution-Grounded Automated AI Research		 arXiv '26 - -
Learn to Discover arXiv
Learning to Discover at Test Time		 arXiv '26 - -
AutoNumerics arXiv
AutoNumerics: An Autonomous, PDE-Agnostic Multi-Agent Pipeline for Scientific Computing		 arXiv '26 - GitHub
SciNav arXiv
SciNav: A General Agent Framework for Scientific Coding Tasks		 arXiv '26 - -
FrontierScience arXiv
FrontierScience: Evaluating AI's Ability to Perform Expert-Level Scientific Tasks		 arXiv '26 - -
EvoDS arXiv
EvoDS: Self-Evolving Autonomous Data Science Agent with Skill Learning and Context Management		 arXiv '26 - -
AutoTTS arXiv
LLMs Improving LLMs: Agentic Discovery for Test-Time Scaling		 arXiv '26 - GitHub
AutoScientists arXiv
AutoScientists: Self-Organizing Agent Teams for Long-Running Scientific Experimentation		 arXiv '26 - -
EurekAgent arXiv
EurekAgent: Agent Environment Engineering is All You Need For Autonomous Scientific Discovery		 arXiv '26 - -

Code Correctness and Reproducibility Assessment

In chronological order, from the earliest to the latest.

Model Paper Venue Website GitHub
Experimental Fidelity arXiv
Beyond Execution: Auditing Experimental Fidelity in LLM-Driven Scientific Research		 arXiv '26 - -
DiscoveryBench arXiv
DiscoveryBench: Towards Data-Driven Discovery with Large Language Models		 arXiv '24 - GitHub
DiscoveryWorld arXiv
DiscoveryWorld: A Virtual Environment for Developing and Evaluating Automated Scientific Discovery Agents		 arXiv '24 - GitHub
InfiAgent-DABench arXiv
InfiAgent-DABench: Evaluating Agents on Data Analysis Tasks		 arXiv '24 - -
ScienceAgentBench arXiv
ScienceAgentBench: Toward Rigorous Assessment of Language Agents for Data-Driven Scientific Discovery		 arXiv '24 - -
LAB-Bench arXiv
Lab-Bench: Measuring Capabilities of Language Models for Biology Research		 arXiv '24 - GitHub
KernelBench arXiv
KernelBench: Can LLMs Write Efficient GPU Kernels?		 arXiv '25 - GitHub
TritonBench arXiv
TritonBench: Benchmarking Large Language Model Capabilities for Generating Triton Operators		 arXiv '25 - GitHub
AstaBench arXiv
AstaBench: Rigorous Benchmarking of AI Agents with a Scientific Research Suite		 arXiv '25 - GitHub
ResearchClawBench arXiv
Probing Scientific General Intelligence of LLMs with Scientist-Aligned Workflows		 arXiv '25 - GitHub
EXP-Bench Website
EXP-Bench: Can AI Conduct AI Research Experiments?		 ICLR '26 - GitHub
PostTrainBench arXiv
PostTrainBench: Can LLM Agents Automate LLM Post-Training?		 arXiv '26 - GitHub
MLReplicate arXiv
MLReplicate: Benchmarking Autonomous Research Systems for Machine Learning Reproducibility		 arXiv '26 - -
BeyondSWE arXiv
BeyondSWE: Can Current Code Agent Survive Beyond Single-Repo Bug Fixing?		 arXiv '26 Website GitHub
NatureBench arXiv
NatureBench: Can Coding Agents Match the Published SOTA of Nature-Family Papers?		 arXiv '26 - -
SciCoQA arXiv
SciCoQA: Quality Assurance for Scientific Paper--Code Alignment		 ACL '26 Website GitHub

4. Tables & Figures

Scientific Figure Generation

In chronological order, from the earliest to the latest.

Model Paper Venue Website GitHub
ChartGPT arXiv
ChartGPT: Leveraging LLMs to Generate Charts from Abstract Natural Language		 arXiv '23 - -
MatPlotAgent arXiv
MatPlotAgent: Method and Evaluation for LLM-Based Agentic Scientific Data Visualization		 arXiv '24 - -
CoDA arXiv
CoDA: Agentic Systems for Collaborative Data Visualization		 arXiv '25 - -
PlotGen arXiv
PlotGen: Multi-Agent LLM-based Scientific Data Visualization via Multimodal Feedback		 arXiv '25 - -
VIS-Shepherd arXiv
VIS-Shepherd: Constructing Critic for LLM-based Data Visualization Generation		 arXiv '25 - -
DiagramAgent arXiv
From Words to Structured Visuals: A Benchmark and Framework for Text-to-Diagram Generation and Editing		 CVPR '25 - -
StarVector arXiv
StarVector: Generating Scalable Vector Graphics Code from Images and Text		 CVPR '25 - -
VisCoder arXiv
VisCoder: Fine-Tuning LLMs for Executable Python Visualization Code Generation		 EMNLP '25 - -
AI-Generated Figures arXiv
AI-Generated Figures		 arXiv '26 - -
AutoFigure-Edit arXiv
AutoFigure-Edit: Generating Editable Scientific Illustration		 arXiv '26 - GitHub
AutoFigure arXiv
AutoFigure-Edit: Generating Editable Scientific Illustration		 ICLR '26 - GitHub
PaperBanana arXiv
PaperBanana: Automating Academic Illustration for AI Scientists		 arXiv '26 - -
SAIL arXiv
Setting SAIL: Leveraging Scientist-AI-Loops for Rigorous Visualization Tools		 arXiv '26 - -
Crafter arXiv
Crafter: A Multi-Agent Harness for Editable Scientific Figure Generation from Diverse Inputs		 arXiv '26 - -
DiagramRAG arXiv
DiagramRAG: A Lightweight Framework to Retrieve Scientific Diagram for Figure Generation		 arXiv '26 - -
GeoSVG-RL arXiv
GeoSVG-RL: Geometry-Aware Reinforcement Learning for Layout-Constrained Text-to-SVG Diagram Generation		 arXiv '26 - -
Can AI Draw Sci. arXiv
Can AI Draw Science? A Benchmark for Evaluating Scientific Figure Generation by Text-to-Image and Multimodal Models		 arXiv '26 - -
SciDiagramEdit arXiv
SciDiagramEdit: Learning to Edit Scientific Diagrams from Paper Revisions		 arXiv '26 - -
GenGA arXiv
GenGA: Editable and Data-Grounded Graphical Abstract Generation for Academic Papers		 arXiv '26 - -

Table Understanding & Generation

In chronological order, from the earliest to the latest.

Model Paper Venue Website GitHub
ArxivDIGESTables arXiv
ArxivDIGESTables: Synthesizing Scientific Literature into Tables using Language Models		 EMNLP '24 - -
Chain-of-Table arXiv
Chain-of-Table: Evolving Tables in Reasoning Chain for Table Understanding		 ICLR '24 - -
ShowTable arXiv
ShowTable: Unlocking Creative Table Visualization with Collaborative Reflection and Refinement		 CVPR '26 - -
Table2LaTeX-RL arXiv
Table2LaTeX-RL: Converting Table Images to High-Fidelity LaTeX Code Using Reinforced Multimodal Language Models		 arXiv '25 - -
CSPO arXiv
CSPO: Alleviating Reward Ambiguity for Structured Table-to-LaTeX Generation		 arXiv '26 - -

Mathematical Formulas & TikZ

In chronological order, from the earliest to the latest.

Model Paper Venue Website GitHub
AutomaTikZ arXiv
AutomaTikZ: Text-Guided Synthesis of Scientific Vector Graphics with TikZ		 ICLR '24 - -
DeTikZify arXiv
DeTikZify: Synthesizing Graphics Programs for Scientific Figures and Sketches with TikZ		 NeurIPS '24 - -
TikZilla arXiv
TikZilla: Scaling Text-to-TikZ with High-Quality Data and Reinforcement Learning		 arXiv '26 - -
Edit2TikZ arXiv
Edit2TikZ: A Comprehensive and Challenging Benchmark for Scientific Figure Editing with TikZ		 arXiv '26 - -

Visual Fidelity and Scientific Accuracy Assessment

In chronological order, from the earliest to the latest.

Model Paper Venue Website GitHub
StructEval Website
StructEval: Benchmarking LLMs' Capabilities to Generate Structural Outputs		 TMLR '25 Website GitHub
PlotCraft arXiv
PlotCraft: Pushing the Limits of LLMs for Complex and Interactive Data Visualization		 arXiv '25 - -
TeXpert Website
TeXpert: Multi-Level Benchmark for LaTeX Code Generation		 SDP '25 - -
AbGen arXiv
AbGen: Evaluating Large Language Models in Ablation Study Design and Evaluation for Scientific Research		 ACL '25 - -
SciFig arXiv
SciFig: Towards Automating Scientific Figure Generation		 arXiv '26 - -
SciFlow-Bench arXiv
SciFlow-Bench: Evaluating Structure-Aware Scientific Diagram Generation via Inverse Parsing		 arXiv '26 - -
FigureBench arXiv
AutoFigure: Generating and Refining Publication-Ready Scientific Illustrations		 ICLR '26 - GitHub
SciFigQual-Bench arXiv
SciFigQual-Bench: A Benchmark for Scientific Figure Quality Assessment with Full-Manuscript Context		 arXiv '26 - -
SciFigAlign arXiv
SciFigAlign: Scoring Scientific Figures by Fine-tuned Alignment of Visuals with Manuscript Evidence		 arXiv '26 - -
SciFigPlag-Bench arXiv
SciFigPlag-Bench: A Benchmark for Provenance-Aware Scientific Figure Plagiarism Detection		 arXiv '26 - -
VLM Blind/Misled arXiv
How Do VLMs Behave When Blind or Misled? Behavioral Evaluation of VLMs on Scientific Figures		 arXiv '26 - -

5. Paper Writing

Semi-Automated Writing Assistance

In chronological order, from the earliest to the latest.

Model Paper Venue Website GitHub
CoAuthor arXiv
CoAuthor: Human-AI Collaborative Writing with Language Models		 arXiv '22 - -
AI Writing Study arXiv
AI Writing Study		 AIED '25 - -
DraftMarks arXiv
DraftMarks: Enhancing Transparency in Human-AI Co-Writing Through Interactive Skeuomorphic Process Traces		 arXiv '25 - -
PaperDebugger arXiv
PaperDebugger: A Plugin-Based Multi-Agent System for In-Editor Academic Writing, Review, and Editing		 arXiv '25 - GitHub
ScholarCopilot arXiv
ScholarCopilot: Training LLMs for Academic Writing with Integrated Citation		 arXiv '25 - -
XtraGPT arXiv
XtraGPT: Context-Aware and Controllable Academic Paper Revision		 arXiv '25 - -
LimAgents arXiv
Multi-Agent LLMs for Generating Research Limitations		 arXiv '26 - -
PaperMentor arXiv
PaperMentor: A Human-Centered Multi-Agent Writing Tutor for AI Research Papers on Overleaf		 arXiv '26 - -
AutoSupervision arXiv
AutoSupervision: Closing the Feedback Loop in Scientific Workflows with Grounded Revision Verification		 arXiv '26 - -
ReasFlow arXiv
ReasFlow: Assisting Reasoning-Centric Scientific Discovery in Applied Mathematics via a Knowledge-Based Multi-Agent System		 arXiv '26 - -

Fully Automated Paper Generation

In chronological order, from the earliest to the latest.

Model Paper Venue Website GitHub
CycleResearcher arXiv
CycleResearcher: Improving Automated Research via Automated Review		 ICLR '25 - -
Agent Laboratory Website
Agent Laboratory: Using LLM Agents as Research Assistants		 EMNLP '25 - -
FutureGen arXiv
FutureGen: A RAG-based Approach to Generate the Future Work of Scientific Article		 arXiv '25 - -
AI Scientist arXiv
The AI Scientist: Towards Fully Automated Open-Ended Scientific Discovery		 Nature '26 - GitHub
APRES arXiv
APRES: An Agentic Paper Revision and Evaluation System		 arXiv '26 - -
LECTOR arXiv
LECTOR: Joint Optimization of Scientific Reasoning Graphs and Introduction Generation		 arXiv '26 - -
RWGBench arXiv
RWGBench: Evaluating Scholarly Positioning in Related Work Generation		 arXiv '26 - -

Societal Analysis

In chronological order, from the earliest to the latest.

Model Paper Venue Website GitHub
AI Writing Adoption Website
AI Writing Adoption		 Nature '26 - -
Nature AI Survey Website
More than Half of Researchers Now Use AI for Peer Review		 Nature '26 - -
Denial of Science arXiv
Distributed Denial of Science: How Indirect Data Poisoning of AI Systems Can Industrialize Scientific Fraud		 arXiv '26 - -
AI Slop OSS arXiv
"AI Slop is DDoSing Open Source": Understanding the Impact of AI-Generated Contributions on Open Source Sustainability		 arXiv '26 - -

Writing Quality and AI Detection Assessment

In chronological order, from the earliest to the latest.

Model Paper Venue Website GitHub
Mapping LLM Use arXiv
Mapping the Increasing Use of LLMs in Scientific Papers		 arXiv '24 - -
CycleReviewer arXiv
CycleResearcher: Improving Automated Research via Automated Review		 ICLR '25 - -
Stanford Agentic Website
Stanford Agentic		 Web '25 - -
SciIG arXiv
Let's Use ChatGPT To Write Our Paper! Benchmarking LLMs To Write the Introduction of a Research Paper		 arXiv '25 - -
Watermarking arXiv
Detecting LLM-Generated Peer Reviews		 arXiv '25 - -
PaperWritingBench arXiv
PaperOrchestra: A Multi-Agent Framework for Automated AI Research Paper Writing		 arXiv '26 - -
CiteTracer arXiv
Source or It Didn't Happen: A Multi-Agent Framework for Citation Hallucination Detection		 arXiv '26 - -
Process Eval arXiv
Process-Oriented Evaluation of AI-Assisted Scientific Writing		 arXiv '26 - -

6. Peer Review

Automated Review Generation

In chronological order, from the earliest to the latest.

Model Paper Venue Website GitHub
ChatReviewer Website
ChatReviewer: ChatGPT-based Paper Reviewing and Response Generation		 GitHub '23 - GitHub
AI-Peer-Review Website
AI-Peer-Review		 GitHub '24 - GitHub
MARG arXiv
MARG: Multi-Agent Review Generation for Scientific Papers		 arXiv '24 - -
Reviewer2 arXiv
Reviewer2: Optimizing Review Generation Through Prompt Generation		 arXiv '24 - -
ReviewRL Website
ReviewRL: Towards Automated Scientific Review with RL		 EMNLP '25 - -
DeepReviewer arXiv
DeepReview: Improving LLM-based Paper Review with Human-like Deep Thinking Process		 arXiv '25 - -
OpenReviewer Website
OpenReviewer: A Specialized Large Language Model for Generating Critical Scientific Paper Reviews		 NAACL '25 - -
REMOR arXiv
REMOR: Automated Peer Review Generation with LLM Reasoning and Multi-Objective Reinforcement Learning		 arXiv '25 - -
ScholarPeer arXiv
ScholarPeer: A Context-Aware Multi-Agent Framework for Automated Peer Review		 arXiv '26 - -
ProReviewer arXiv
From Passive Generation to Investigation: A Proactive Scientific Peer Review Agent		 arXiv '26 - -
PeerCheck arXiv
PeerCheck: Enhancing LLM-Generated Academic Reviews Towards Human-Level Quality		 arXiv '26 - -
Local Pre-Screening arXiv
Local AI pre-screening for human triple-blind peer review in health sciences		 arXiv '26 - -
ReVoicer arXiv
ReVoicer: Conversational Voice Annotation for Human-Centered, LLM-Assisted Peer Review		 arXiv '26 - -

Meta-Review & Reviewer Matching

In chronological order, from the earliest to the latest.

Model Paper Venue Website GitHub
Metag arXiv
Metag: A dataset to build agentic meta-reviewing capabilities		 arXiv '26 - -
AgentReview Website
AgentReview: Exploring Peer Review Dynamics with LLM Agents		 EMNLP '24 - -
Meta-Review LLMs Website
Meta-Review LLMs		 NAACL '25 - -
RATE arXiv
RATE: Reviewer Profiling and Annotation-free Training for Expertise Ranking in Peer Review Systems		 arXiv '26 - -
MERIT arXiv
MERIT: Matching Expertise via Rubric-Informed Training for Reviewer Assignment		 arXiv '26 - -

Adversarial Attacks & Bias Analysis

In chronological order, from the earliest to the latest.

Model Paper Venue Website GitHub
Raina etal arXiv
Raina etal		 EMNLP '24 - -
AI Review Lottery arXiv
The AI Review Lottery: Widespread AI-Assisted Peer Reviews Boost Paper Scores and Acceptance Rates		 arXiv '24 - -
Ye etal arXiv
Ye etal		 arXiv '24 - -
Breaking the Reviewer arXiv
Breaking the Reviewer: Assessing the Vulnerability of Large Language Models in Automated Peer Review Under Textual Adversarial Attacks		 arXiv '25 - -
LLM Reviewer Bias arXiv
LLM Reviewer Bias		 arXiv '25 - -
Prompt Injection arXiv
Prompt Injection Attacks on LLM Generated Reviews of Scientific Publications		 arXiv '25 - -
Sahoo etal arXiv
Sahoo etal		 arXiv '25 - -
Zhou etal arXiv
Zhou etal		 arXiv '25 - -
Presentation Gaming arXiv
No Hidden Prompts Needed! You Can Game AI Peer Review with Presentation-Only Revisions		 arXiv '26 - -
LLMs Favor LLMs? arXiv
Do LLMs Favor LLMs? Quantifying Interaction Effects in Peer Review		 arXiv '26 - -
Gaming AI Reviews arXiv
Gaming AI-Assisted Peer Reviews Poses New Risks to the Scientific Community		 arXiv '26 - -
Phantom Refs arXiv
Phantom References: Hallucinated Citations That Survive Peer Review at Top-Tier Conferences		 arXiv '26 - -
Rhetorical Reward-Hacking arXiv
How Can Rhetoric Reward-Hack AI Reviewers? Dissecting Rhetorical Sensitivity in AI-Based Peer Review		 arXiv '26 - -

Detection & Policy

In chronological order, from the earliest to the latest.

Model Paper Venue Website GitHub
AI Detection arXiv
Is Your Paper Being Reviewed by an LLM? Benchmarking AI Text Detection in Peer Review		 arXiv '25 - -
AI Use Rejects Website
Major Conference Catches Illicit AI Use — and Rejects Hundreds of Papers		 Nature '26 - -
Nature AI Survey Website
More than Half of Researchers Now Use AI for Peer Review		 Nature '26 - -
Policy Enforcement arXiv
Policy Enforcement		 arXiv '26 - -
Reviewer Feedback Website
What Happens When Reviewers Receive AI Feedback in Their Reviews?		 CHI '26 - -
AAAI-26 Pilot arXiv
AI-Assisted Peer Review at Scale: The AAAI-26 AI Review Pilot		 arXiv '26 - -
Reviewer AI Policies arXiv
AI-Assisted Peer Review Across Research Communities: From Reviewer AI Policies to LLM Review Quality		 arXiv '26 - -

Review Consistency and Bias Assessment

In chronological order, from the earliest to the latest.

Model Paper Venue Website GitHub
VERA-RL arXiv
Not Just Reason, Not Just Scan: Reinforcement Learning for Proactive Scientific Error Verification over Academic Paper		 EMNLP '26 - -
Review Survey Website
More than Half of Researchers Now Use AI for Peer Review — often Against Guidance		 IF '25 - -
Stanford Agentic Website
Stanford Agentic		 Web '25 - -
ClaimCheck Website
ClaimCheck: How Grounded are LLM Critiques of Scientific Papers?		 EMNLP '25 - -
REFUTE Website
REFUTE: A Benchmark for Scientific Critique and Epistemic Calibration in Language Models		 HF '26 Website -
ReViewGraph arXiv
Automatic Paper Reviewing with Heterogeneous Graph Reasoning over LLM-Simulated Reviewer-Author Debates		 AAAI '26 - -
ReviewAgents arXiv
ReviewAgents: Bridging the Gap Between Human and AI-Generated Paper Reviews		 arXiv '25 - -
ICLR 2025 Study Website
ICLR 2025 Study		 NMI '26 - -
AI Reviewer Limits arXiv
On the limits and opportunities of AI reviewers: Reviewing the reviews of Nature-family papers with 45 expert scientists		 arXiv '26 - -
PRISM arXiv
PRISM: A Multi-Dimensional Benchmark for Evaluating LLM Peer Reviewers		 arXiv '26 - -
LLM-Human Alignment arXiv
How Closely Do LLM Reviews Align with Human Peer Review?		 arXiv '26 - -
Epistemic Reliability arXiv
Articulate Intuition or Genuine Analysis? Benchmarking Epistemic Reliability in LLM-as-a-Judge Peer Reviews		 arXiv '26 - -
SurveyReview arXiv
SurveyReview: A Reviewer-Aligned Benchmark for Survey Evaluators		 arXiv '26 - -

7. Rebuttal

Reviewer Comment Analysis

In chronological order, from the earliest to the latest.

Model Paper Venue Website GitHub
ReviewMT arXiv
Peer Review as A Multi-Turn and Long-Context Dialogue with Role-Based Interactions		 arXiv '24 - -
ICLR Rebuttal Study arXiv
ICLR Rebuttal Study		 arXiv '25 - -
RbtAct arXiv
RbtAct: Rebuttal as Supervision for Actionable Review Feedback Generation		 arXiv '26 - -
GoodPoint arXiv
GoodPoint: Learning Constructive Scientific Paper Feedback from Author Responses		 arXiv '26 - -

Automated Rebuttal Generation

In chronological order, from the earliest to the latest.

Model Paper Venue Website GitHub
ReviewerToo arXiv
ReviewerToo: Should AI Join The Program Committee? A Look At The Future of Peer Review		 arXiv '25 - -
RebuttalAgent arXiv
RebuttalAgent: Strategic Persuasion in Academic Rebuttal via Theory of Mind		 ICLR '26 - GitHub
Author-in-the-Loop arXiv
Author-in-the-Loop Response Generation and Evaluation: Integrating Author Expertise and Intent in Responses to Peer Review		 ACL '26 - -
DRPG arXiv
DRPG: An Agentic Framework for Academic Rebuttal		 arXiv '26 - GitHub
Paper2Rebuttal arXiv
Paper2Rebuttal: A Multi-Agent Framework for Transparent Author Response Assistance		 arXiv '26 - -
Defend arXiv
Defend: Automated Rebuttals for Peer Review with Minimal Author Guidance		 arXiv '26 - -

Rebuttal Effectiveness Assessment

In chronological order, from the earliest to the latest.

Model Paper Venue Website GitHub
Re$^2$ arXiv
Re$^2$		 arXiv '25 - -
Commitment Checklist arXiv
Commitment Checklist: Auditing Author Commitments in Peer Review		 arXiv '26 - -
Re$^3$Align arXiv
Re$^3$Align		 ACL '26 - -
Rebuttals Move arXiv
Rebuttals Move Peer-Review Scores, but Initial-Review Structure Bounds the Movement		 arXiv '26 - -
Trust AI Reviews arXiv
To Trust or Not to Trust: Authors' Response to AI-based Reviews		 arXiv '26 - -

8. Dissemination (Paper2X)

Paper2Poster

In chronological order, from the earliest to the latest.

Model Paper Venue Website GitHub
P2P Website
P2P: Automated Paper-to-Poster Generation and Fine-Grained Benchmark		 ICLR '26 - -
Paper2Poster Website
Paper2Poster: Towards Multimodal Poster Automation from Scientific Papers		 NeurIPS '25 - GitHub
PosterForest arXiv
PosterForest: Hierarchical Multi-Agent Collaboration for Scientific Poster Generation		 arXiv '25 - -
PosterGen arXiv
PosterGen: Aesthetic-Aware Paper-to-Poster Generation via Multi-Agent LLMs		 arXiv '25 - -
APEX arXiv
APEX: Academic Poster Editing Agentic Expert		 arXiv '26 - GitHub
PosterOmni arXiv
PosterOmni: Generalized Artistic Poster Creation via Task Distillation and Unified Reward Feedback		 arXiv '26 - -
Any2Poster arXiv
Any2Poster: Any-Source Poster Generation Across Modalities and Domains		 arXiv '26 - -
PosterMELD arXiv
PosterMELD: Multi-Agent Paper-to-Poster Generation for Controllable Design Diversity with Editable Print-Ready Outputs		 arXiv '26 - -

Paper2Slides

In chronological order, from the earliest to the latest.

Model Paper Venue Website GitHub
DOC2PPT Website
DOC2PPT: Automatic Presentation Slides Generation from Scientific Documents		 AAAI '22 - -
PPTAgent arXiv
PPTAgent: Generating and Evaluating Presentations Beyond Text-to-Slides		 EMNLP '25 - GitHub
AutoPresent arXiv
AutoPresent: Designing Structured Visuals from Scratch		 CVPR '25 - -
Paper2Slides Website
Paper2Slides: From Paper to Presentation in One Click		 GitHub '25 - GitHub
Auto-Slides arXiv
Auto-Slides: An Interactive Multi-Agent System for Creating and Customizing Research Presentations		 arXiv '25 - -
PASS arXiv
PASS: Presentation Automation for Slide Generation and Speech		 arXiv '25 - -
SlideGen arXiv
SlideGen: Collaborative Multimodal Agents for Scientific Slide Generation		 arXiv '25 - -
Talk to Your Slides arXiv
Talk to Your Slides: Efficient Slide Editing Agent		 arXiv '25 - -
SlideTailor arXiv
SlideTailor: Personalized Presentation Slide Generation for Scientific Papers		 AAAI '26 - GitHub
DeepPresenter arXiv
DeepPresenter: Environment-Grounded Reflection for Agentic Presentation Generation		 arXiv '26 - GitHub
Office Raccoon Website
Office Raccoon		 Web '26 - -
X+Slides arXiv
X+Slides: Benchmarking Audience-Conditioned Slide Generation		 arXiv '26 - -
SeaSlides arXiv
SeaSlides: Semantic Abstraction Layer for Agentic Slide Generation		 arXiv '26 - -

Paper2Video

In chronological order, from the earliest to the latest.

Model Paper Venue Website GitHub
Preacher Website
Preacher: Paper-to-Video Agentic System		 ICCV '25 - GitHub
Paper2Video arXiv
Paper2Video: Automatic Video Generation from Scientific Papers		 arXiv '25 - GitHub
PresentAgent Website
PresentAgent: Multimodal Agent for Presentation Video Generation		 EMNLP '25 - GitHub
PresentAgent-2 arXiv
PresentAgent-2: Towards Generalist Multimodal Presentation Agents		 arXiv '26 - -
Paper2Video Talks arXiv
A Good Talk Does not Look Like a Summary, It Teaches You! Measuring Takeaways from Paper-to-Video Talks		 arXiv '26 - -

Paper2Web & Social Media

In chronological order, from the earliest to the latest.

Model Paper Venue Website GitHub
Paper2Web arXiv
Paper2Web: Let's Make Your Paper Alive!		 arXiv '25 - GitHub
ResearchStudio-Reel arXiv
ResearchStudio-Reel: Automate the Last Mile of Research from Paper to Poster, Video, and Blog		 arXiv '26 - -
I-WebGenBench arXiv
I-WebGenBench: Evaluating Interactivity in LLM-Generated Scientific Web Applications		 arXiv '26 - -
SciForge arXiv
SciForge: An AI-Native, Multimodal Workbench for Scientific Discovery		 arXiv '26 - -

Fidelity and Adoption Assessment

In chronological order, from the earliest to the latest.

Model Paper Venue Website GitHub
PPTEval arXiv
PPTAgent: Generating and Evaluating Presentations Beyond Text-to-Slides		 EMNLP '25 - GitHub
PresentQuiz arXiv
Paper2Video: Automatic Video Generation from Scientific Papers		 arXiv '25 - GitHub
PresentEval Website
PresentAgent: Multimodal Agent for Presentation Video Generation		 EMNLP '25 - GitHub
Sci. Comm. Correspondence arXiv
Unifying Scientific Communication: Fine-Grained Correspondence Across Scientific Media		 arXiv '26 - -

9. End-to-End Systems

Fully Automated Research Systems

In chronological order, from the earliest to the latest.

Model Paper Venue Website GitHub
Gemini Co-Scientist arXiv
Accelerating Scientific Research with Gemini in the Real-World		 arXiv 2026 - -
AutoResearch arXiv
AutoResearch: Insight In, Hallucination Out		 arXiv 2026 - -
Eureka arXiv
Eureka: Task-Conditioned Meta-Agent Orchestration for Scientific Discovery		 arXiv 2026 - -
AutoResearch Diagnosis arXiv
How Do Agents Fail on AutoResearch: End-to-End Diagnostic Evaluation on 100 Real-World Frontier Research Tasks		 arXiv 2026 - -
ResearchTown arXiv
ResearchTown: Simulator of Human Research Community		 ICML 2025 Website GitHub
Agent Laboratory arXiv
Agent Laboratory: Using LLM Agents as Research Assistants		 arXiv 2025 - -
AgentRxiv arXiv
AgentRxiv: Towards Collaborative Autonomous Research		 arXiv 2025 - -
ARIS - GitHub 2025 - GitHub
freephdlabor arXiv
Build Your Personalized Research Group: A Multiagent Framework for Continual and Interactive Science Automation		 arXiv 2025 - -
SciMaster arXiv
SciMaster: Towards General-Purpose Scientific AI Agents		 arXiv 2025 - GitHub
- arXiv
Towards End-to-End Automation of AI Research		 Nature 2026 Website GitHub
Idea2Story arXiv
Idea2Story: An Automated Pipeline for Transforming Research Concepts into Complete Scientific Narratives		 arXiv 2026 - -
UniScientist - Web 2026 - -
ASI-Evolve - GitHub 2026 - GitHub
FARS - Web 2026 - -
AutoResearchClaw - GitHub 2026 - GitHub
CORAL arXiv
CORAL: Towards Autonomous Multi-Agent Evolution for Open-Ended Discovery		 arXiv 2026 - GitHub
AutoSOTA arXiv
AutoSOTA: An End-to-End Automated Research System for State-of-the-Art AI Model Discovery		 arXiv 2026 - GitHub
AiScientist-LH arXiv
Toward Autonomous Long-Horizon Engineering for ML Research		 arXiv 2026 - GitHub
OpenResearcher (2026) arXiv
OpenResearcher: A Fully Open Pipeline for Long-Horizon Deep Research Trajectory Synthesis		 arXiv 2026 - GitHub
Aletheia arXiv
Towards Autonomous Mathematics Research		 arXiv 2026 - GitHub
AutoSci arXiv
AutoSci: A Memory-Centric Agentic System for the Full Scientific Research Lifecycle		 arXiv 2026 - -
ScientistOne arXiv
ScientistOne: Towards Human-Level Autonomous Research via Chain-of-Evidence		 arXiv 2026 - -
Arbor arXiv
Toward Generalist Autonomous Research via Hypothesis-Tree Refinement		 arXiv 2026 - -
Agon arXiv
Agon: An Autonomous Large-Scale Omnidisciplinary Research System Built on Prompt Economy		 arXiv 2026 - GitHub
ResearchClawBench arXiv
ResearchClawBench: A Benchmark for End-to-End Autonomous Scientific Research		 arXiv 2026 - -
Act as Real Researcher arXiv
Act As a Real Researcher: A Suite of Benchmarks Evaluating Frontier LLMs and Agentic Harnesses in Research Lifecycle		 arXiv 2026 - -
ResearchArena arXiv
How Far Are We From True Auto-Research?		 arXiv 2026 - -
NVAITC AI Sci. arXiv
NVAITC AI Scientist: A Governed End-to-End Research System -- A Hypertension GWAS Case Study		 arXiv '26 - -
DeepResearch Agent Sys. arXiv
DeepResearch Agent System		 arXiv '26 - -
ScienceFlow arXiv
ScienceFlow: A long-horizon agent for ML research, scientific discovery and beyond		 arXiv '26 - -
Spark-to-Paper arXiv
Spark-to-Paper: End-to-End Research Paper Generation as a Composable Skill		 arXiv '26 - -
OmniScientist arXiv
OmniScientist: An Omni-Modal Omni-Discipline AI Scientist		 arXiv '26 - -

Domain-Specific Systems

In chronological order, from the earliest to the latest.

Model Paper Venue Website GitHub
AutoSR arXiv
AutoSR: Automatic Symbolic Regression by Searching Research States		 arXiv 2026 - -
The AI Engineer arXiv
Closed-loop AI achieves certifiable engineering design		 arXiv 2026 - -
AlphaFold 3 arXiv
Accurate Structure Prediction of Biomolecular Interactions with AlphaFold 3		 Nature 2024 Website -
Medical AI Scientist arXiv
Towards a Medical AI Scientist		 arXiv 2026 - -
Cognitive Scientist arXiv
Closing the Loop to Discover Psychological Theories with an Automated Cognitive Scientist		 arXiv '26 - -
Molecular Closed-Loop arXiv
Closed-loop Auto Research for Molecular Property Prediction: Discovering and Certifying Generalizable Improvements		 arXiv '26 - -
Auto Research (Materials) arXiv
Auto Research for Materials: Auditable AI-Scientist Workflows with Held-Out Transfer		 arXiv '26 - -
OmniQEC arXiv
OmniQEC: discovering practical quantum error-correcting codes by an AI scientist		 arXiv '26 - -
Andy arXiv
Andy: A Mathematical Agent for Rigorous Proof and Autonomous Research		 arXiv '26 - -
BaT arXiv
BaT: Towards Self-Evolving Medical Research Agent with Stage Rubrics		 arXiv '26 - -
Agentic Sci. Community arXiv
An Agentic AI Scientific Community for Automated Neural Operator Discovery		 arXiv '26 - -
WARA arXiv
WARA: A Closed-Loop Multi-Agent Framework for Wireless Optimization Autoresearch		 ICCC '26 - -

Evolutionary & Self-Improving Systems

In chronological order, from the earliest to the latest.

Model Paper Venue Website GitHub
ShinkaEvolve arXiv
ShinkaEvolve: Towards Open-Ended and Sample-Efficient Program Evolution		 arXiv 2025 - GitHub
Darwin Godel Machine arXiv
Darwin Godel Machine: Open-Ended Evolution of Self-Improving Agents		 arXiv 2025 - GitHub
EvoMaster arXiv
EvoMaster: A Foundational Evolving Agent Framework for Agentic Science at Scale		 arXiv 2026 - -
Self-Driving Lab arXiv
Compressing the Validation Bottleneck: An Agentic Self-Driving Lab for Scientific Discovery		 arXiv '26 - -
Meta-Reflection arXiv
Autonomous Scientific Discovery via Iterative Meta-Reflection		 arXiv '26 - -
Project2Task arXiv
Project2Task: Graph-Guided Project-Level Planning for Autonomous Research		 arXiv '26 - -
Adaptive Inquiry arXiv
Beyond Solution-Centric Search: Adaptive Inquiry and Knowledge Revision for Autonomous ML Engineering		 arXiv '26 - -
CausalForge arXiv
CausalForge: A Formally Grounded, Self-Improving Agentic Framework for Automated Research in Causal Inference		 arXiv '26 - -
World-Model Scaling arXiv
Scaling Automatic Research Agents via World Models		 arXiv '26 - -
Research Preference Models arXiv
AI Research Preference Models		 arXiv '26 - -

Research Platforms & Infrastructure

In chronological order, from the earliest to the latest.

Model Paper Venue Website GitHub
Symposium arXiv
Symposium: Trust via Auditable Records for Communities of AI Scientist Agents		 arXiv 2026 - -
Towards an AI co-scientist arXiv
Towards an AI co-scientist		 arXiv 2025 - -
PiFlow arXiv
PiFlow: Principle-aware Scientific Discovery with Multi-Agent Collaboration		 arXiv 2025 - -
LabClaw - Web 2026 - -
- arXiv
OpenAI Is Throwing Everything into Building a Fully Automated Researcher		 MIT TR 2026 Website -
XScientist arXiv
XScientist: A Git-Like Research Protocol for Long-Running Autonomous Scientific Discovery		 arXiv '26 - -
Auditable Records arXiv
From Trajectories to Evidence: Auditable Experimental Records for Industrial Research Agents		 arXiv '26 - -
PEARL arXiv
PEARL: Auditable Repair for Scientific Reasoning Graph Extraction		 arXiv '26 - -
IdeaTrail arXiv
IdeaTrail: Full-Process Agent Trajectories for Scientific Ideation		 arXiv '26 - -
EviGraph arXiv
EviGraph: Evidence-Guided Autonomous Research Agents		 arXiv '26 - -
Discovery Environments arXiv
Scaling Scientific Discovery Environments for Turn-Level Agentic RL		 arXiv '26 - -
Quantum Theorem Bench arXiv
Benchmarking Agents for Proving Theorems in Quantum Algorithms and Quantum Information		 arXiv '26 - -
SciDiscovery Readiness arXiv
Are LLMs Ready for Scientific Discovery? A Capability-Oriented Benchmark for AI Scientists		 arXiv '26 - -
ARAC arXiv
ARAC: Benchmarking Auto-Research's Alignment and Completeness on End-to-End Researchs		 arXiv '26 - -
Beyond Final Scores arXiv
Beyond Final Scores: A Systematic Evaluation of Agents for Long-Horizon AI Research and Development		 arXiv '26 - -
Replication Training arXiv
Training AI Scientists to Replicate Research		 arXiv '26 - -
Rubric Meta-Eval arXiv
Can LLMs Write Reliable Rubrics? A Meta-Evaluation for Experiment Reproduction		 arXiv '26 - -
Fast-Moving Domains arXiv
Adversarial Fast-Moving Real-World Domains as Test Beds for Benchmarking AI Scientist Capabilities		 arXiv '26 - -
AutoWorldModel-Bench arXiv
AutoWorldModel-Bench: A State-Centric Benchmark for Automated World-Model Research		 arXiv '26 - -
Scientific Memory arXiv
Beyond Memory Leaderboards: Evaluating Scientific Memory as Budgeted Context Restoration		 arXiv '26 - -
SciHazard arXiv
SciHazard: A Benchmark for Measuring Scientific Safety Risks with Decomposed Harm Scoring		 arXiv '26 - -
Science Edge Eval. arXiv
Science Edge Evaluation: SEE the Missing Step Toward Real Scientific Discovery		 arXiv '26 - -

10. Societal & Critical Perspectives

⏲️ In chronological order, from the earliest to the latest.

Model Paper Venue Website GitHub
- arXiv
Science Done on a Machine by a Machine: AI Agents in Computational Chemistry		 arXiv 2026 - -
- arXiv
Bayesian Optimization for Self-Driving Materials Laboratories: From Algorithms to Physics-Informed Workflows		 arXiv 2026 - -
- arXiv
LLM assisted writing deserves empirical evaluation		 arXiv 2026 - -
- arXiv
A Metric with Positive Sectional Curvature on $S^2\times S^3$ 		arXiv 2026 - -
- arXiv
Weak-Type Bounds for Convolution on the Boolean Hypercube		 arXiv 2026 - -
- arXiv
Quo Vadis? Scientific Discovery in the Age of Artificial Intelligence		 Theory of Science - -
- arXiv
Are LLMs becoming similarly creative? Evidence from three years of models		 arXiv 2026 - -
- arXiv
Navigating the Jagged Technological Frontier		 Org. Sci. 2025 Website -
- arXiv
Reassessing Academic Integrity in the Age of AI		 SSH Open 2025 Website -
The AI Deskilling Paradox arXiv
The AI Deskilling Paradox		 CACM 2025 Website -
Hidden Pitfalls of AI Scientist Systems arXiv
The More You Automate, the Less You See: Hidden Pitfalls of AI Scientist Systems		 arXiv 2025 - -
Rethinking Science in the Age of AI arXiv
Rethinking Science in the Age of Artificial Intelligence		 arXiv 2025 - -
- arXiv
Measuring AI Ability to Complete Long Tasks		 METR 2025 Website -
- arXiv
Towards a Science of Scaling Agent Systems		 arXiv 2025 - -
- arXiv
Artificial Intelligence Tools Expand Scientists' Impact but Contract Science's Focus		 Nature 2026 Website -
- [arXiv](https://www.cell.com/patterns/fulltext/S2666-3899(25)
AI for Scientific Discovery is a Social Problem		 Patterns 2026 [Website](https://www.cell.com/patterns/fulltext/S2666-3899(25) -
Research Integrity in the Age of AI arXiv
Research Integrity and Academic Authority in the Age of Artificial Intelligence: From Discovery to Curation?		 arXiv 2026 - -
SciSciGPT arXiv
SciSciGPT: Advancing Human-AI Collaboration in the Science of Science		 Nature CS 2026 Website -
SimStep arXiv
SimStep: Chain-of-Abstractions for Incremental Specification and Debugging of AI-Generated Interactive Simulations		 arXiv 2025 - -
ConvoLearn arXiv
ConvoLearn: A Learning Sciences Grounded Dataset for Fine-Tuning Dialogic AI Tutors		 arXiv 2026 - -
AFIM: Academic Fraud Inclination Metric arXiv
AFIM: Academic Fraud Inclination Metric		 Web 2026 Website -
- arXiv
AI Researchers' Views on Automating AI R&D and Intelligence Explosions		 arXiv 2026 - -
- arXiv
AI Scientists Are Changing Research		 Nature 2026 Website -
Learning by Creating (Talk) arXiv
Learning by Creating: A Human-Centered Vision for AI in Education		 Talk 2026 Website -
AI Sci. No Reasoning arXiv
AI scientists produce results without reasoning scientifically		 arXiv 2026 - -
Not Built for Autonomy arXiv
Agentic AI Scientists Are Not Built For Autonomous Scientific Discovery		 arXiv 2026 - -
Workflow != Closure arXiv
Workflow Closure Is Not Scientific Closure in Auto-Research Systems		 arXiv 2026 - -
PseudoBench arXiv
PseudoBench: Measuring How Agentic Auto-Research Fuels Pseudoscience		 arXiv 2026 - -
SciIntegrity-Bench arXiv
SciIntegrity-Bench: A Benchmark for Evaluating Academic Integrity in AI Scientist Systems		 arXiv 2026 - -
- arXiv
Abduction Without a Body? Representational Grounding and the Abduction Loop for Scientific Hypothesis Generation		 arXiv '26 - -
- arXiv
Visualizing Graph-to-Answer Mechanism Recovery in Materials-Science Hypothesis Generation		 arXiv '26 - -
- arXiv
Generative AI use in Statistical Research: A Literature Review and Code Generation Case Study		 arXiv '26 - -
- arXiv
AI's Capability in Assisting Scientific Research in Physics, Astrophysics, and Cosmology I: Literature Review		 arXiv '26 - -
- arXiv
Scientific Knowledge Discovery in the Age of Large Language Models		 arXiv '26 - -
- arXiv
Is Deep Research Reliable? Misleading Knowledge Induces False Conclusions		 arXiv '26 - -
- arXiv
Diagnosing Search Behavior and Failure Modes in Long-Horizon Search Agents		 arXiv '26 - -
- arXiv
A Pathway to General-Purpose Scientific AI: Multimodal Comprehension of Scientific Images		 arXiv '26 - -
- arXiv
Position: AI Agents in Scientific Teams Should Be Studied as Human-Agent Systems		 arXiv '26 - -
- arXiv
The Past and Future of AI Scientists		 arXiv '26 - -
- arXiv
Efficiency Matters in Autonomous Research		 arXiv '26 - -
- arXiv
Agentic Auto-Research is Fuzz Testing		 arXiv '26 - -
- arXiv
Mechanist: AI as a Scientific Instrument for Discovering the Mechanisms of Intelligence		 arXiv '26 - -
- arXiv
An AI Scientist that Doesn't Drift: Taste, Structure, and Falsifiable Findings in a Quadruped Navigation Research Loop		 arXiv '26 - -
- arXiv
Long-Horizon Autonomous Architecture Research with a Language-Model Agent: A Behavioural Case Study		 arXiv '26 - -
- arXiv
Stress-testing large language model agents in a robotic chemistry laboratory		 arXiv '26 - -

11. Surveys & Curated Lists

⏲️ In chronological order, from the earliest to the latest.

Model Paper Venue Website GitHub
LLM4SR arXiv
LLM4SR: A Survey on Large Language Models for Scientific Research		 arXiv 2025 - -
From Automation to Autonomy arXiv
From Automation to Autonomy: A Survey on Large Language Models for Scientific Discovery		 arXiv 2025 - -
AI4Research arXiv
AI4Research: A Survey of Artificial Intelligence for Scientific Research		 arXiv 2025 - -
A Survey of AI Scientists arXiv
A Survey of AI Scientists		 arXiv 2025 - -
- arXiv
Large Language Models for Scientific Idea Generation: A Creativity-Centered Survey		 arXiv 2025 - -
- arXiv
Large Language Models for Automated Scholarly Paper Review: A Survey		 Inf. Fusion 2025 Website -
AutoResearch Survey Website
What's Missing in Autonomous Research? A Systematization of Systems, Benchmarks, and Verification		 Web 2026 Website -

12. Tools & GitHub Repos

Open-source tools, frameworks, and curated resource lists for AI-assisted research (not directly tied to a single paper).

Curated Lists

Repository Stars Description
Awesome-Deep-Research GitHub Up-to-date collection of agentic deep research resources
Awesome-Scientific-Language-Models GitHub Survey of scientific LLMs (EMNLP'24)
Awesome-LLM-Scientific-Discovery GitHub Three-level autonomy framework (EMNLP'25)
Awesome-AI-Scientist-Papers GitHub Resources on AI Scientist systems
Awesome-Auto-Research-Tools GitHub Automated research tools catalog
awesome-autoresearch GitHub Autonomous improvement loops and research agents
awesome-ai-research-writing GitHub Prompt templates and agent skills for AI-assisted writing

Idea Generation

Repository Stars Description
Virtual-Scientists GitHub VirSci: multi-agent collaborative idea generation (ACL'25)
ResearchAgent GitHub Iterative idea proposal with reviewing agents

Literature Review

Repository Stars Description
paper-qa GitHub PaperQA2: superhuman RAG for scientific Q&A
local-deep-research GitHub Fully local deep research
researchgpt GitHub Conversational interaction with research papers
gpt-researcher GitHub Autonomous agent for comprehensive online research
AutoSurvey GitHub Automated comprehensive literature surveys
storm GitHub Wikipedia-style article generation (STORM)
lune-mcp-server GitHub MCP client for the hosted Lune Research corpus: paper search, citation traversal, evidence extraction, quote-backed claim verification (API key required)

Coding & Experiments

Repository Stars Description
autoresearch (Karpathy) GitHub Autonomous ML experiments, ~12 exp/hour overnight
Paper2Code GitHub Multi-agent ML paper to code transformation
RD-Agent GitHub Microsoft's LLM framework for autonomous data science
MLAgentBench GitHub 13 end-to-end ML experimentation tasks
SWE-bench GitHub Real-world GitHub issue resolution benchmark
Thoth GitHub Dashboard-first Claude Code and Codex runtime for durable autoresearch runs, work-item locks, ledgers, and reviewable verdicts

Peer Review

Repository Stars Description
paper-reviewer GitHub arXiv paper reviews + blog posts
ai-peer-review GitHub Multi-LLM reviews + meta-review synthesis
openreviewer GitHub Llama-8B fine-tuned on 79K expert reviews

⬆ Back to Top

Last updated: 2026-08-31 · Maintained by WorldBench

About

🔥 A Survey on AI Auto-Research

worldbench.github.io/awesome-ai-auto-research

Topics

agentagentic-aiaiai-researchai-scientistauto-researchlarge-language-modelsllmmulti-agent-systemrag

Resources

Readme
MIT license
Activity
Custom properties

Stars

538 stars

Watchers

16 watching

Forks

41 forks
Report repository

Releases

Packages

Contributors

Languages