AutoResearchBench is a new benchmark showing top AI agents achieve under 10% success on complex scientific literature discovery tasks that demand deep comprehension and open-ended search.
hub Canonical reference
Tongyi DeepResearch Technical Report
Canonical reference. 77% of citing Pith papers cite this work as background.
52
Pith papers citing it
Background
77% of classified citations
abstract
We present Tongyi DeepResearch, an agentic large language model, which is specifically designed for long-horizon, deep information-seeking research tasks. To incentivize autonomous deep research agency, Tongyi DeepResearch is developed through an end-to-end training framework that combines agentic mid-training and agentic post-training, enabling scalable reasoning and information seeking across complex tasks. We design a highly scalable data synthesis pipeline that is fully automatic, without relying on costly human annotation, and empowers all training stages. By constructing customized environments for each stage, our system enables stable and consistent interactions throughout. Tongyi DeepResearch, featuring 30.5 billion total parameters, with only 3.3 billion activated per token, achieves state-of-the-art performance across a range of agentic deep research benchmarks, including Humanity's Last Exam, BrowseComp, BrowseComp-ZH, WebWalkerQA, xbench-DeepSearch, FRAMES and xbench-DeepSearch-2510. We open-source the model, framework, and complete solutions to empower the community.
hub tools
citation-role summary
background 10
baseline 2
method 1
citation-polarity summary
representative citing papers
ChartWalker provides a hierarchical knowledge graph construction method and structure-aware sampling to generate cross-chart RAG benchmarks, releasing ChartWalker-Bench that exposes performance gaps across RAG paradigms.
SAGE-OPD improves multi-turn OPD via turn-level selective intervention, teacher-confidence weighting, and loss normalization, reporting up to 13.3% relative gain in ALFWorld unseen success rate over standard OPD.
PhySciBench benchmark shows current AI models achieve at most 33.5% accuracy on physical science tasks; DelveAgent framework improves accuracy by up to 7.5 points and cuts costs to one-third.
EvoBrowseComp is an auto-updatable benchmark of 800 complex questions for search agents, synthesized via a three-agent live-web framework to ensure temporal freshness and block parametric shortcuts.
FORT synthesizes shortcut-resistant search tasks by controlling four identified shortcut risks across entity selection, graph construction, question formulation, and refinement, producing training data that yields agents with longer search trajectories and top performance among open-source models on
PixelRAG shows that operating RAG entirely over web screenshots outperforms text-based retrieval on NQ, SimpleQA, MMSearch, LiveVQA, and MoNaCo, with up to 18.1% accuracy gains and 3x token savings via image compression.
VibeSearchBench provides 200 tasks across 20 domains with progressive-disclosure simulation and graph-matching evaluation, showing frontier LLM agents achieve at most 30.30 F1 on long-horizon proactive search.
REFLECT benchmark shows current LLM judges achieve below 55% accuracy detecting failures in evidence-based research agents, especially on evidence verification.
ClawForge is a generator framework that creates reproducible executable benchmarks for command-line agents under state conflict, with ClawForge-Bench showing frontier models reach at most 45.3% strict accuracy and that state inspection drives most performance gaps.
ActGuide-RL uses human action data as plan-style guidance in mixed-policy RL to overcome exploration barriers in LLM agents, matching SFT+RL performance on search benchmarks without cold-start training.
CuSearch reallocates rollout budget in RLVR toward deeper-search trajectories as a proxy for retrieval supervision density, yielding up to 11.8 exact-match gains over uniform GRPO sampling on ZeroSearch.
AIDA is the first end-to-end autonomous agent that combines a domain-specific language with Pareto-guided reinforcement learning to discover insights from complex business data.
Starling, a multi-agent LLM system, extracts ~6.3 million nuanced structured records from PubMed across six tasks with reported error rates of 0.6-7.7%, lower than several curated databases.
Small 7B reasoning models were fine-tuned on synthetic and curated QFT problems using RL and SFT, yielding performance gains, error analysis, and public release of data and traces.
Mind-ParaWorld creates parallel worlds with atomic facts to evaluate search agents on future scenarios, showing they synthesize evidence well but struggle with collection, coverage, sufficiency judgment, and stopping decisions.
SAGE with MHFA improves failure recovery in autonomous research agents, raising metrics-bearing outputs from 42% to 92% on a 12-topic benchmark versus single-reflection baselines.
ICBCBench is a new consortium-built benchmark that jointly measures retrieval-reasoning accuracy and end-to-end report quality for deep research agents in finance.
Agents-K1 is an end-to-end pipeline with a multimodal parser, 4B GRPO-trained extractor, and agent CLI that builds scientific knowledge graphs from full papers and was run on 2.46 million documents to produce Scholar-KG.
Arbor combines a coordinator, executors, and a hypothesis tree to enable cumulative autonomous research, outperforming Codex and Claude Code by over 2.5x on six real tasks and reaching 86.36% Any Medal on MLE-Bench Lite.
DAC decomposes agentic search into cooperative searcher and generator agents with cross-agent signals (abstention reward and hard-positive augmentation), achieving strong QA benchmark performance via LoRA on a shared backbone.
Deep research agents exhibit widespread search-time contamination on six public benchmarks, with three defined leakage types inflating performance by up to 4%.
Harness-1 uses a state-externalizing harness for RL-trained search agents and reports 0.730 average curated recall, outperforming the next open subagent by 11.4 points.
DR-rubric is a two-stage framework using iterative agentic search to generate atomic verifiable constraints for GRPO-based RL, achieving competitive performance on 6 benchmarks with 1K-3K examples via bootstrap or frontier-model rubrics.
citing papers explorer
-
ChartWalker: Benchmarking the Cross-Chart RAG Task with Hierarchical Knowledge Graphs
ChartWalker provides a hierarchical knowledge graph construction method and structure-aware sampling to generate cross-chart RAG benchmarks, releasing ChartWalker-Bench that exposes performance gaps across RAG paradigms.
-
SAGE-OPD: Selective Agent-Guided Intervention for Multi-Turn On-Policy Distillation
SAGE-OPD improves multi-turn OPD via turn-level selective intervention, teacher-confidence weighting, and loss normalization, reporting up to 13.3% relative gain in ALFWorld unseen success rate over standard OPD.
-
Deep Research in Physical Sciences: A Multi-Agent Framework and Comprehensive Benchmark
PhySciBench benchmark shows current AI models achieve at most 33.5% accuracy on physical science tasks; DelveAgent framework improves accuracy by up to 7.5 points and cuts costs to one-third.
-
EvoBrowseComp: Benchmarking Search Agents on Evolving Knowledge
EvoBrowseComp is an auto-updatable benchmark of 800 complex questions for search agents, synthesized via a three-agent live-web framework to ensure temporal freshness and block parametric shortcuts.
-
FORT-Searcher: Synthesizing Shortcut-Resistant Search Tasks for Training Deep Search Agents
FORT synthesizes shortcut-resistant search tasks by controlling four identified shortcut risks across entity selection, graph construction, question formulation, and refinement, producing training data that yields agents with longer search trajectories and top performance among open-source models on
-
PIXELRAG: Web Screenshots Beat Text for Retrieval-Augmented Generation
PixelRAG shows that operating RAG entirely over web screenshots outperforms text-based retrieval on NQ, SimpleQA, MMSearch, LiveVQA, and MoNaCo, with up to 18.1% accuracy gains and 3x token savings via image compression.
-
VibeSearchBench: Benchmarking Long-horizon Proactive Search in the Wild
VibeSearchBench provides 200 tasks across 20 domains with progressive-disclosure simulation and graph-matching evaluation, showing frontier LLM agents achieve at most 30.30 F1 on long-horizon proactive search.
-
Time to REFLECT: Can We Trust LLM Judges for Evidence-based Research Agents?
REFLECT benchmark shows current LLM judges achieve below 55% accuracy detecting failures in evidence-based research agents, especially on evidence verification.
-
ClawForge: Generating Executable Interactive Benchmarks for Command-Line Agents
ClawForge is a generator framework that creates reproducible executable benchmarks for command-line agents under state conflict, with ClawForge-Bench showing frontier models reach at most 45.3% strict accuracy and that state inspection drives most performance gaps.
-
Learning Agentic Policy from Action Guidance
ActGuide-RL uses human action data as plan-style guidance in mixed-policy RL to overcome exploration barriers in LLM agents, matching SFT+RL performance on search benchmarks without cold-start training.
-
CuSearch: Curriculum Rollout Sampling via Search Depth for Agentic RAG
CuSearch reallocates rollout budget in RLVR toward deeper-search trajectories as a proxy for retrieval supervision density, yielding up to 11.8 exact-match gains over uniform GRPO sampling on ZeroSearch.
-
Towards Autonomous Business Intelligence via Data-to-Insight Discovery Agent
AIDA is the first end-to-end autonomous agent that combines a domain-specific language with Pareto-guided reinforcement learning to discover insights from complex business data.
-
Fine-Tuning Small Reasoning Models for Quantum Field Theory
Small 7B reasoning models were fine-tuned on synthetic and curated QFT problems using RL and SFT, yielding performance gains, error analysis, and public release of data and traces.
-
Evaluating the Search Agent in a Parallel World
Mind-ParaWorld creates parallel worlds with atomic facts to evaluate search agents on future scenarios, showing they synthesize evidence well but struggle with collection, coverage, sufficiency judgment, and stopping decisions.
-
One Reflection Is Not Enough: Self-Correcting Autonomous Research via Multi-Hypothesis Failure Attribution
SAGE with MHFA improves failure recovery in autonomous research agents, raising metrics-bearing outputs from 42% to 92% on a 12-topic benchmark versus single-reflection baselines.
-
ICBCBench: An Industry Consortium Benchmark for Financial Deep Research
ICBCBench is a new consortium-built benchmark that jointly measures retrieval-reasoning accuracy and end-to-end report quality for deep research agents in finance.
-
Agents-K1: Towards Agent-native Knowledge Orchestration
Agents-K1 is an end-to-end pipeline with a multimodal parser, 4B GRPO-trained extractor, and agent CLI that builds scientific knowledge graphs from full papers and was run on 2.46 million documents to produce Scholar-KG.
-
Toward Generalist Autonomous Research via Hypothesis-Tree Refinement
Arbor combines a coordinator, executors, and a hypothesis tree to enable cumulative autonomous research, outperforming Codex and Claude Code by over 2.5x on six real tasks and reaching 86.36% Any Medal on MLE-Bench Lite.
-
Divide and Cooperate: Role-Decomposed Multi-Agent LLM Training with Cross-Agent Learning Signals
DAC decomposes agentic search into cooperative searcher and generator agents with cross-agent signals (abstention reward and hard-positive augmentation), achieving strong QA benchmark performance via LoRA on a shared backbone.
-
Search-Time Contamination in Deep Research Agents: Measuring Performance Inflation in Public Benchmark Evaluation
Deep research agents exhibit widespread search-time contamination on six public benchmarks, with three defined leakage types inflating performance by up to 4%.
-
Harness-1: Reinforcement Learning for Search Agents with State-Externalizing Harnesses
Harness-1 uses a state-externalizing harness for RL-trained search agents and reports 0.730 average curated recall, outperforming the next open subagent by 11.4 points.
-
Deep Research as Rubric for Reinforcement Learning
DR-rubric is a two-stage framework using iterative agentic search to generate atomic verifiable constraints for GRPO-based RL, achieving competitive performance on 6 benchmarks with 1K-3K examples via bootstrap or frontier-model rubrics.
-
Enhancing LLM Metacognition via Cognitive Pairwise Training
CPT is introduced as a pairwise reasoning-trace comparison stage that improves the reasoning-metacognition trade-off over standard SFT+RL pipelines across model scales.
-
Beyond Trajectory Rewards: Step-level Credit Assignment for Agentic Search via Graph Modeling
GDCR assigns step-level rewards via distance to the answer node in a training-time ER graph and SAPO combines these with trajectory advantages for credit assignment in agentic search.
-
AgentFugue: Agent Scaling for Long-Horizon Tasks through Collective Reasoning
AgentFugue introduces a plug-in shared reasoning hub trained with SFT and RL that enables peer agents to share intermediate reasoning, yielding gains on long-horizon tasks over strong baselines.
-
Efficient Agentic Reasoning Through Self-Regulated Simulative Planning
SR²AM achieves competitive Pass@1 accuracy on diverse tasks with 25.8-95.3% fewer reasoning tokens than much larger models by using self-regulated simulative planning trained via supervised learning and RL.
-
Argus: Evidence Assembly for Scalable Deep Research Agents
Argus coordinates a Navigator and multiple Searchers via an evidence graph for deep research, reporting average gains of 5.5 points with one Searcher and 12.7 points with eight parallel Searchers across eight benchmarks, reaching 86.2 on BrowseComp with 64 Searchers.
-
ToolCUA: Towards Optimal GUI-Tool Path Orchestration for Computer Use Agents
ToolCUA introduces a trajectory scaling pipeline and staged RL to optimize GUI-tool switching, reaching 46.85% accuracy on OSWorld-MCP for a 66% relative gain over baseline.
-
HTPO: Towards Exploration-Exploitation Balanced Policy Optimization via Hierarchical Token-level Objective Control
HTPO introduces hierarchical token-level objective control in RLVR to balance exploration and exploitation by grouping tokens according to difficulty, correctness, and entropy, yielding up to 8.6% gains on AIME benchmarks over DAPO.
-
LongSeeker: Elastic Context Orchestration for Long-Horizon Search Agents
Context-ReAct enables agents to dynamically manage context via five atomic operations, and LongSeeker fine-tuned on 10k trajectories achieves 61.5% and 62.5% on BrowseComp benchmarks, outperforming prior agents.
-
Towards Knowledgeable Deep Research: Framework and Benchmark
The paper introduces the KDR task, HKA multi-agent framework, and KDR-Bench to enable LLM agents to integrate structured knowledge into deep research reports, with experiments showing outperformance over prior agents.
-
TimelineReasoner: Advancing Timeline Summarization with Large Reasoning Models
TimelineReasoner applies large reasoning models in a Global Cognition plus Detail Exploration loop to produce more accurate, complete, and coherent timelines from news than prior LLM-based methods.
-
MiroThinker: Pushing the Performance Boundaries of Open-Source Research Agents via Model, Context, and Interactive Scaling
MiroThinker shows that scaling agent-environment interactions via reinforcement learning lets a 72B open-source model reach up to 81.9% on GAIA and approach commercial performance on research benchmarks.
-
Yuvion LLM: An Adversarially-Aware Large Language Model for Content And AI Safety
Yuvion LLM applies adversarially aware training and introduces the YLRE benchmark set, claiming superior safety robustness over larger models on multiple tasks.
-
PBSD: Privileged Bayesian Self-Distillation for Long-Horizon Credit Assignment
PBSD derives autoregressive turn-level credit signals from outcome rewards via the posterior-to-prior ratio converted through Bayes' rule between student and privileged teacher models.
-
Struct-Searcher: Agentic Structural Thinking Advances Multimodal Deep Information Seeking
Struct-Searcher introduces a structural agentic workflow grounded in belief revision theory that maintains an evolving multimodal graph for conflict-aware deep information seeking and reports accuracy gains on several VL benchmarks.
-
Rethinking Continual Experience Internalization for Self-Evolving LLM Agents
Existing methods for turning LLM interaction experience into parametric skills collapse over multiple iterations; principle-level experience, step-wise injection, and off-policy teacher distillation yield more stable continual learning.
-
Scaling Retrieval-Augmented Reasoning with Parallel Search and Explicit Merging
MultiSearch uses parallel multi-query retrieval plus explicit merging inside a reinforcement-learning loop to improve retrieval-augmented reasoning, outperforming baselines on seven QA benchmarks.
-
ViDR: Grounding Multimodal Deep Research Reports in Source Visual Evidence
ViDR treats source figures as retrievable and verifiable evidence objects in multimodal deep research reports and introduces MMR Bench+ to measure improvements in visual integration and verifiability.
-
Towards On-Policy Data Evolution for Visual-Native Multimodal Deep Search Agents
Proposes image-bank harness and ODE closed-loop data generation to boost multimodal deep search agents, reporting average score gains from 24.9% to 39.0% on 8 benchmarks for 8B model and 30.6% to 41.5% for 30B.
-
Mind DeepResearch Technical Report
MindDR combines a Planning Agent, DeepSearch Agent, and Report Agent with SFT cold-start, Search-RL, Report-RL, and preference alignment to reach competitive scores on research benchmarks using 30B-scale models.
-
Reward Hacking in the Era of Large Models: Mechanisms, Emergent Misalignment, Challenges
The paper introduces the Proxy Compression Hypothesis as a unifying framework explaining reward hacking in RLHF as an emergent result of compressing high-dimensional human objectives into proxy reward signals under optimization pressure.
-
SimpleSearch-VL: A Simple Recipe for Multimodal Agentic Deep Search
SimpleSearch-VL improves Qwen3-VL multimodal agent baselines by 15.8-16 points on average using 7K total training examples and reaches parity with Gemini-3-Pro on the 30B variant.
-
BioInsight: Multi-Agent Orchestration for Interactive Biomedical Knowledge Discovery
BioInsight is a multi-agent system that generates interactive, provenance-preserving biomedical evidence interfaces from disease names and protein data.
-
CAREAgent: Clinical Agent with Structured Reasoning and Tool-Integrated for Order Generation
CAREAgent uses two-stage data construction, supervised fine-tuning, and reinforcement learning to improve clinical order generation F1 scores on benchmarks including unseen ClinicalBench.
-
Valley3: Scaling Omni Foundation Models for E-commerce
Valley3 is an omni MLLM for e-commerce that uses a four-stage pre-training pipeline plus post-training for controllable reasoning and agentic search, outperforming baselines on e-commerce benchmarks while staying competitive on general ones.
-
MedXIAOHE: A Comprehensive Recipe for Building Medical MLLMs
MedXIAOHE is a medical MLLM that claims state-of-the-art benchmark performance through specialized pretraining to cover long-tail diseases and RL-based reasoning training.