AgentFlare

Scholar · tool use and function calling in language models

Tool Use in Language Models: Research Digest

10 foundational papers with 11,129 combined citations, plus the latest arXiv preprints.

Foundational papers1011,129 citations
Most citedToolformer: Language models can teach…8,208 citations · 2023
arXiv, Sep 2026102 papers+96% vs Oct 2025
Newest preprint2026-10-27

AISA Scholar API · arXiv

arXiv submissions per monthPapers matching (abs:"tool use" OR abs:"function calling") AND abs:"language model" · last 12 complete months
monthly arXiv submissions: Oct 2025 52 → Sep 2026 102; high 103 (May 2026), low 41 (Nov 2025)366084108Oct 2025Mar 2026Sep 2026
402 · Pay-per-Crawl$0.02

You're reading the full report — free for humans and search engines. The same content ships as a structured JSON dataset (normalized facts, sources and citation arrays, no HTML parsing) so AI agents can consume it directly — metered per fetch via the open HTTP 402 payment protocol.

⚡ Agents: fetch .data.json · $0.02 → agents: crawlerAutoPrice=true · verify at https://esa.aisa.one/api/v1/access/verify

What’s new on arXiv

A12026-10-27cs.LG

Scaling Laws for Tool-Use Agents via Controllable Environment Synthesis

Andrew Szot, Samira Abnar, Omar Attia, Abbas Kazerouni, Michael Kirchhof, Cheng-Yu Hsieh, Raviteja Vemulapalli, Alexander Toshev

We explore how to build performant large language model (LLM) tool-use agents that complete complex, multi-step tasks. Specifically, we study how data composition in synthetically generated tool-use environments affects the training of such agents. To do so, we propose the SynthAPI pipeline, which explicitly controls the diversity of generated environments along the axes of tools, tasks, and the databases on which the tools operate. By systematically varying SynthAPI's generation controls in 458,752 tasks across 849 trained models of four different sizes, we derive a scaling law that predicts downstream tool-use agent generalization based on the scale and composition of the synthesized environments. These predictable relationships reveal several lessons for scaling tool-use agent data. As data budgets grow, scaling tool and database diversity becomes more effective than increasing the number of tasks, and increasing data complexity yields larger gains. Scaling according to these findings yields agents that outperform models trained on data from prior work in large-scale tool-use agent and synthetic environment generation across four evaluation benchmarks.

A22026-10-27cs.RO

Which Hand Was Late? Diagnosis for Robot Failure Recovery in Bimanual Tool Use

Baptiste Bonin, Caro Strickland, Audrey Durand

Failure diagnosers for robot manipulation are increasingly vision--language models (VLMs), trained and evaluated on single-arm failures produced by perturbing a successful demonstration. Bimanual tool use adds failures of other kinds: a tool that slips out of its socket while still held, two correct gestures made at different times, and attempts that fail for two reasons at once. On a simulated bimanual tool-use task whose failure causes are read from the recorded trace, we measure what each sensing channel lets a diagnoser tell. The tool's slip is read on the tool hand; which hand was late is read on both arms, while a sensor on one hand reads its own hand's delay and at most part of the other's; and whether a late attempt was also pushed too weakly is not identifiable from passive sensing when the tool hand is late. Within two attempts, a repair policy that escalates when the cause is not identifiable repairs as many failures as one told the true cause. A VLM's internal state carries the late hand almost perfectly, along a direction no closer to the one its answer reads than a random one; whether the answer reports it depends on the wording of the question.

A32026-10-27cs.CV

PerceptRL: Making Visual Perception Rewardable in MLLMs from Answers to Interactions

Tongkun Guan, Zhibo Yang, Mingkun Yang, Haocheng Wang, Wei Shen, Xiaokang Yang

Fine-grained visual perception remains a bottleneck during post-training of multimodal large language models (MLLMs), as perceptual errors can constrain gains from reasoning-focused optimization. Although image-caption pre-training improves perception, its influence can weaken as subsequent stages change data mixtures and objectives. Continuing caption supervision during post-training is challenging because subjective, information-dense descriptions complicate verifiable reward design. We present PerceptRL, which strengthens perception at two complementary levels. Perception RL uses a multi-agent pipeline to extract verified errors from dense image captions and convert the corresponding visual facts into QA pairs with fixed answer labels and rule-based rewards. Chain QA links correlated errors, while multiple-choice QA combines unrelated errors. A specialized route targets high-error domains with tailored verification, and a domain-agnostic general route covers errors beyond these domains. Perception Agentic RL addresses fine-grained and dense cases unresolved by single-turn Perception RL through multi-turn inspection with visual tools. We observe coordinate drift even in models trained to use tools, where incorrect crops or annotations distort evidence and propagate errors. Coordinate Injection corrects spatial operations in cold-start trajectories, while a grounding-aware process reward supervises intermediate spatial fidelity alongside final-answer correctness. Training proceeds through shared cold-start SFT, single-turn Perception RL, and multi-turn Perception Agentic RL. Across seven benchmarks under a shared tool-use protocol, PerceptRL achieves 71.6% average accuracy, surpassing the base model with the same tools and the strongest evaluated frontier VLM by 15.7 and 5.2 percentage points, respectively.

A42026-10-27cs.LG

CoHARL: Co-Evolving Agent Policies and Executable Harnesses with Reinforcement Learning

Jianwen Chen, Peng Xia, Kaide Zeng, Siwei Han, Zhen-Yu Zhang, Charles Fleming, Xinyu Ye, Huaxiu Yao

Large language model (LLM) agents interact with their environments through executable harnesses that govern tool use and feedback. Optimizing a harness for a fixed policy overlooks how its behavior changes during learning, while training a policy under a fixed harness leaves the interface that generates its training experience unchanged. This motivates adapting the harness throughout policy training. We introduce CoHARL, a population-based reinforcement learning framework that co-evolves a shared policy and a population of executable harnesses. The policy learns from rollouts across the population, and a proposer uses its execution traces to revise harness code during training. CoHARL achieves a 27.6% mean resolved rate on SWE-rebench, improving over harness-only search by 6.4 percentage points. Experiments on SWE-bench Verified and Terminal-Bench 2.0 show gains in software engineering and terminal-based tasks. Under matched training budgets, CoHARL also outperforms searching for a harness once before RL and alternating policy updates with revisions to a single active harness.

A52026-10-25cs.AI

HobbyLM: A Sparse Mixture-of-Experts Model Family on a Hobby Budget

Harish Santhanalakshmi Ganesan, Hariharan M. S., Prabhurajhan R

We describe HobbyLM, a family of small sparse Mixture-of-Experts (MoE) models trained from scratch with a few hundred to a few thousand dollars of rented GPU time. The core is a decoder-only transformer with fine-grained experts, sigmoid routing with auxiliary-loss-free load balancing, grouped-query attention with per-head QK-normalization, and aMuon/AdamW optimizer split. Each of these choices was selected with a single-variable ablation suite at 130M parameters. On this architecture we train a 500M-parameter model (169M active) on 40B FineWeb tokens and post-train it into a chat model, a masked-diffusion language model, a function-calling and computer-use agent, and a single "omni" checkpoint that reads text, images, video, sound and speech through frozen encoders and small projectors. We run these models on a laptop CPU with a from-scratch Rust inference engine. Two of the authors then adopted the architecture without architectural changes to build HobbyLM-1B (1.04B total, 305M active parameters), trained on a 100B-token curated mixture with a high-quality anneal. HobbyLM-1B reaches a 47.5 average on seven zero-shot benchmarks, and its builders also release an instruction-tuned model and an on-device macOS computer-use adapter. Finally, we describe HobbyLM-Image, a 333M-parameter flow-matching diffusion transformer that generates 1024x1024 images in a 32x-compressed latent space, together with a dependency-light CPU implementation of the full text-to-image pipeline. We report what worked, what did not, and where the small-scale ceilings lie. We also state for each component which author designed or built it.

A62026-10-25cs.CR

SentryBench: A Cost-Sensitive Benchmark for Adversarial Robustness of Tool-Using LLM Agents in Public-Safety Deployments

Md Abu Yusuf, Oyshi Tabassum Aditi, Md Faysal Hosain Mazed

Large language model (LLM) agents are increasingly wired to consequential tools that send public alerts, dispatch responders, and query citizen records, where one erroneous tool call carries real-world cost. Existing agent-safety benchmarks measure generic harm or prompt-injection success and implicitly treat refusing as the safe default. In public-safety deployments that assumption breaks: an agent too cautious to dispatch help fails as surely as one that triggers a false alert. We present SentryBench, which scores agents on a two-sided, cost-sensitive frontier jointly measuring attack success, over-refusal (missing a real emergency), and over-reaction (unwarranted action) under adversarial tool use, with scenarios grounded in real emergency-call transcripts, disaster messages, and real injection/jailbreak payloads. We formalize deployment harm as an asymmetric cost functional, prove naive caution is harm-dominated whenever it refuses more genuine emergencies than the attack harm it averts, and prove a targeted trust gate adds zero over-refusal on clean input. Across 840 scored episodes over three frontier models and three defenses, refuse-more caution drives over-refusal to 100 percent on two of three models for at most a 3.7-point change in attack success and yields the worst harm everywhere, whereas our trust-gating wrapper dominates the frontier (lowest harm on every model, over-refusal at or below baseline, higher containment). Provider content filters also silently perform part of the defending and must be reported separately.

A72026-10-25cs.AI

Task Success Is Not Enough: Side-effect-Aware Evaluation of Tool-Using Language Model Agents

Jiaju Huang, Shaobin Chen, Xinglong Liang, Xinyu Ma, Hao Yang, Wei Ke, Tao Tan, Yue Sun

Tool-using language model agents are usually evaluated by whether they finish the user's task. That metric misses a common failure: an agent can complete the requested work while also taking an unauthorized side-effecting action, such as writing to the wrong object, expanding the scope of the request, or contacting an unintended recipient. We call this failure mode collateral damage (CD). We introduce SABench (Side-effect-Aware Benchmark), a benchmark of 201 email, calendar, and ticket tasks. Each task crosses one of four state structures (independent, dependent, externally reachable, cross-tool) with one of four trap families (ambiguity, mis-targeting, cascade harm, overreach), and each task has a deterministic trace-based oracle for both goal satisfaction and authorization compliance. Across 2,211 model-task pairs from 11 frontier LLMs spanning major commercial and open model families, 15.2% of all runs incur CD, with risk concentrated in ambiguity tasks (38.8% violation rate) while non-ambiguous traps average 8.4% CD. Among goal-satisfied runs, 12.8% still produce unauthorized writes. Cross-tool and dependent states have about twice the CD rate of independent states. CD rates vary by 4.7x across models at comparable goal rates, which suggests that task success alone is a poor proxy for safe tool use. The SABench task corpus is released at https://huggingface.co/datasets/Phaedrus/sabench; the traces, oracle, code, and analysis pipeline are released at https://github.com/hjj194/sabench-artifact.

A82026-10-25cs.LG

Self-Evolution Reasoner: Continuous Optimization via Policy-Intrinsic Exploration and Exploitation

Wenhang Shi, Yiren Chen, Zhe Zhao, Jinhao Dong, Wei Lu, Xiaoyong Du

Reinforcement learning with verifiable rewards (RLVR) enables large language models (LLMs) to continuously self-improve their reasoning capabilities. Compared to RLVR's scalar rewards, recent self-supervised on-policy distillation methods leverage denser supervision signals. However, both paradigms suffer from exploration bottlenecks, struggling to escape local optima when the policy fails to sample valid solutions. Relying on external demonstrations can bypass this issue; however, the off-policy nature hinders genuine self-improvement and the mismatched-distribution reasoning can not be fully exploited. In this work, we propose Self-Evolution Policy Optimization (SEPO), a framework that drives continuous reasoning capability evolution via guided exploration and in-distribution exploitation. Specifically, SEPO utilizes labels to elicit correct reasoning from the policy, unlocking the exploration potential. By treating current policy conditioned on this in-distribution reasoning as a self-teacher, SEPO distills the newly discovered solution back into the policy. Through this iterative cycle of policy-intrinsic exploration and exploitation, SEPO achieves robust and continuous capability evolution. Across scientific, tool-use, and complex logical reasoning domains, SEPO consistently achieves the best performance under identical training time or steps, demonstrating superior efficiency and convergence ceilings. Notably, SEPO maintains stable improvements even in scenarios where baselines severely underperform or fail entirely, such as on complex tasks or with small-scale models. Our code is available at https://github.com/Eric8932/SEPO.

Foundational papers

  1. 2023
    Toolformer: Language models can teach themselves to use tools8,208 citations
  2. 2024
    Hammer: Robust function-calling for on-device language models via function masking64 citations
  3. 2024
    Toolllm: Facilitating large language models to master 16000+ real-world apis2,714 citations
  4. 2025
    Meta-tool: Unleash open-world function calling capabilities of general-purpose large language models8 citations
  5. 2025
    Robust function-calling for on-device language model via function masking26 citations
  6. 2025
    Alignment for efficient tool calling of large language models36 citations
  7. 2026
    Looped Language Models Improve Compositional Tool Calling3 citations
  8. 2026
    Beyond function calling: Benchmarking tool-using agents under tool-environment unreliability5 citations
  9. 2026
    The Bitter Lesson of Tool Calling1 citations
  10. 2026-10
    Latest on arXiv: Scaling Laws for Tool-Use Agents via Controllable Environment Synthesiscs.LG
Citations per paperCitation counts via the AISA Scholar API
citations: R1 Toolformer: Language … 8,208, R2 Nexusraven: a commerc… 64, R3 Hammer: Robust functi… 64, R4 Meta-tool: Unleash op… 8, R5 Robust function-calli… 26, R6 Looped Language Model… 3, R7 Toolllm: Facilitating… 2,714, R8 Alignment for efficie… 36, R9 Beyond function calli… 5, R10 The Bitter Lesson of … 1R1 Toolformer: Language …8,208R2 Nexusraven: a commerc…64R3 Hammer: Robust functi…64R4 Meta-tool: Unleash op…8R5 Robust function-calli…26R6 Looped Language Model…3R7 Toolllm: Facilitating…2,714R8 Alignment for efficie…36R9 Beyond function calli…5R10 The Bitter Lesson of …1
R120238,208 citations

Toolformer: Language models can teach themselves to use tools

T Schick, J Dwivedi-Yu, R Dessì et al.

… models excel. In this paper, we show that LMs can teach themselves to use external tools via … We introduce Toolformer, a model trained to decide which APIs to call, when to call them, …

R620263 citations

Looped Language Models Improve Compositional Tool Calling

AC Popescu, HSO Borde, P Liò

… Accuracy on multistep tool use … language models are a promising architecture for agentic systems that require reliable planning, coordination, and execution of compositional tool use …

R8202536 citations

Alignment for efficient tool calling of large language models

H Xu, Z Wang, Z Zhu, L Pan, X Chen et al.

… We also analyze over-tool-reliance, where the model uses tools … -tool-reliance with accuracy, leading to unnecessary computational overhead. Conversely, our method reduces over-tool…

R1020261 citations

The Bitter Lesson of Tool Calling

I Patel, S Sen, E Lumer, VK Subbiah

… language models (LLMs) now act as toolcalling agents, invoking external services through APIs that require the model to emit a structured JSON object at each function call (… For models …

How this page is made. Foundational papers and citation counts come from academic search via the AISA Scholar API, with abstracts read from each paper’s page; recent preprints and monthly volumes come from the public arXiv API (metadata CC0). The digest is written by a language model from those abstracts only — every theme links the papers it draws on, and figures not in the inputs are removed.

Sources & citations

  1. Toolformer: Language models can teach themselves to use tools 8,208 citations
  2. Nexusraven: a commercially-permissive language model for function calling 64 citations
  3. Hammer: Robust function-calling for on-device language models via function masking 64 citations
  4. Meta-tool: Unleash open-world function calling capabilities of general-purpose large language models 8 citations
  5. Robust function-calling for on-device language model via function masking 26 citations
  6. Looped Language Models Improve Compositional Tool Calling 3 citations
  7. Toolllm: Facilitating large language models to master 16000+ real-world apis 2,714 citations
  8. Alignment for efficient tool calling of large language models 36 citations
  9. Beyond function calling: Benchmarking tool-using agents under tool-environment unreliability 5 citations
  10. The Bitter Lesson of Tool Calling 1 citations
  11. Scaling Laws for Tool-Use Agents via Controllable Environment Synthesis arXiv · 2026-10-27
  12. Which Hand Was Late? Diagnosis for Robot Failure Recovery in Bimanual Tool Use arXiv · 2026-10-27
  13. PerceptRL: Making Visual Perception Rewardable in MLLMs from Answers to Interactions arXiv · 2026-10-27
  14. CoHARL: Co-Evolving Agent Policies and Executable Harnesses with Reinforcement Learning arXiv · 2026-10-27
  15. HobbyLM: A Sparse Mixture-of-Experts Model Family on a Hobby Budget arXiv · 2026-10-25
  16. SentryBench: A Cost-Sensitive Benchmark for Adversarial Robustness of Tool-Using LLM Agents in Public-Safety Deployments arXiv · 2026-10-25
  17. Task Success Is Not Enough: Side-effect-Aware Evaluation of Tool-Using Language Model Agents arXiv · 2026-10-25
  18. Self-Evolution Reasoner: Continuous Optimization via Policy-Intrinsic Exploration and Exploitation arXiv · 2026-10-25

Related