Static SFT and RL training for tool-use agents leads to performance drops under open-world distributional shifts across perception, interaction, reasoning and internalization; perturbation-augmented fine-tuning is proposed as mitigation.
hub Mixed citations
arXiv preprint arXiv:2503.23383 , year=
Mixed citation behavior. Most common role is background (67%).
28
Pith papers citing it
Background
67% of classified citations
hub tools
citation-role summary
background 7
baseline 1
other 1
citation-polarity summary
representative citing papers
SENTINEL generates targeted tasks from model failures in a Controller-Proposer-Solver loop, raising Pass^1 from 66.4 to 74.9 on Tau2-Bench Retail and outperforming standard RL.
CARL trains a critic for segment-level credit assignment from binary outcomes in LLM tool-use trajectories, yielding 6.7-9.7 point accuracy gains and 53% fewer calls on solvable questions across five benchmarks.
A normative-descriptive framework shows LLMs' tool-calling perceptions misalign with true need/utility for web search, and hidden-state estimators improve decisions over self-perceived baselines.
ECHO is a selective turn-memory framework for agentic RL that compresses turns into indexed records, selects them for bounded contexts, and uses source indices to assign outcome credit to supporting evidence, reaching 43.4% accuracy on BrowseComp-Plus versus 28.9% for GRPO and 36.1% for SUPO.
APPO refines branching and credit assignment in agentic RL via a Branching Score and procedure-level scaling, improving baselines by nearly 4 points on 13 benchmarks.
AXPO addresses the Thinking-Acting Gap in agentic RL training by targeted resampling of tool calls in all-wrong subgroups, delivering +1.8pp gains over GRPO on nine multimodal benchmarks with an 8B model beating a 32B baseline on Pass@4.
HASP upgrades textual skills into executable Program Functions that intervene in LLM agent loops at inference, post-training, or self-evolution, delivering 25% gains over ReAct and 30.4% over Search-R1 on reasoning benchmarks.
ComplexMCP benchmark shows top LLM agents achieve under 60% success on dynamic interdependent tool tasks versus 90% for humans, due to tool retrieval saturation, over-confidence, and strategic defeatism.
PruneTIR prunes erroneous tool-call trajectories during LLM inference via three trigger-based components to raise Pass@1 accuracy and efficiency while shortening context.
SOD reweights on-policy distillation strength step-by-step using divergence to stabilize tool use in small language model agents, yielding up to 20.86% gains and 26.13% on AIME 2025 for a 0.6B model.
Agent-World autonomously synthesizes verifiable real-world tasks and uses continuous self-evolution to train 8B and 14B agents that outperform proprietary models on 23 benchmarks.
ATTC reduces 'Tool Ignored' errors in tool-integrated reasoning by adaptively trusting tool results according to generated code confidence, yielding 4.1-7.5% gains across models and datasets.
CharTool equips MLLMs with cropping and code tools plus agentic RL on DuoChart data to raise chart-reasoning accuracy by up to 9.78 percent on benchmarks.
MARL-Rad trains region-specific and global agents with reinforcement learning on clinical rewards to produce more accurate radiology reports than prior methods on MIMIC-CXR and IU X-ray datasets.
Strengthening LLM reasoning through RL, SFT, or chain-of-thought prompting increases tool hallucination rates on SimpleToolHalluBench, with a reliability-capability trade-off observed across mitigation attempts.
Survey that defines agentic RL for LLMs via POMDPs, introduces a taxonomy of planning/tool-use/memory/reasoning capabilities and domains, and compiles open environments from over 500 papers.
WebThinker equips large reasoning models with autonomous web exploration and interleaved reasoning-drafting via a Deep Web Explorer and RL-based DPO training, yielding gains on GPQA, GAIA, and report-generation benchmarks.
ReTool uses outcome-driven RL to train 32B LLMs to dynamically use code tools during reasoning, reaching 72.5% accuracy on AIME and surpassing o1-preview.
IAPO is an RL method that aligns model input attributions with a teacher to improve tool-calling in multimodal SLMs, reporting 3% average VQA accuracy gains on Qwen2.5-VL-3B across six tests.
Instance-level experiential knowledge provides strong gains for LLM tool calling, parallel sampling activates it more effectively than deeper reasoning, and RL-based internalization outperforms SFT, yielding the KATE framework with consistent benchmark improvements.
SlimSearcher reduces tool-call rounds by 17-58% on GAIA, BrowseComp and XBenchDeepSearch while maintaining accuracy via Pareto filtration in SFT and Adaptive Reward Gating in RL.
TAO-RL improves agentic RL by filtering degenerate trajectories and reshaping advantages with tool-aware entropy bonuses, yielding better performance on reasoning benchmarks.
E3-TIR integrates expert prefixes, guided branches, and self-exploration via mix policy optimization to deliver 6% better tool-use performance with under 10% of the usual synthetic data and 1.46x ROI.
citing papers explorer
-
ComplexMCP: Evaluation of LLM Agents in Dynamic, Interdependent, and Large-Scale Tool Sandbox
ComplexMCP benchmark shows top LLM agents achieve under 60% success on dynamic interdependent tool tasks versus 90% for humans, due to tool retrieval saturation, over-confidence, and strategic defeatism.
-
Agent-World: Scaling Real-World Environment Synthesis for Evolving General Agent Intelligence
Agent-World autonomously synthesizes verifiable real-world tasks and uses continuous self-evolution to train 8B and 14B agents that outperform proprietary models on 23 benchmarks.
-
The Landscape of Agentic Reinforcement Learning for LLMs: A Survey
Survey that defines agentic RL for LLMs via POMDPs, introduces a taxonomy of planning/tool-use/memory/reasoning capabilities and domains, and compiles open environments from over 500 papers.
-
UI-TARS-2 Technical Report: Advancing GUI Agent with Multi-Turn Reinforcement Learning
UI-TARS-2 reaches 88.2 on Online-Mind2Web, 47.5 on OSWorld, 50.6 on WindowsAgentArena, and 73.3 on AndroidWorld while attaining 59.8 mean normalized score on a 15-game suite through multi-turn RL and scalable data generation.