Research map / Large language models / Instruction and fine-tuning
ToolACE-R: Model-aware Iterative Training and Adaptive Refinement for Tool learning
Xingshan Zeng, Weiwen Liu, Xu Huang, Zezhong Wang, Lingzhi Wang, Liangyou Li, Yasheng Wang, Lifeng Shang, Xin Jiang, Ruiming Tang, Qun Liu
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| remove model-aware data selection (ablation) | ToolACE-R · BFCL Overall | accuracy: 86.33 -> 85.04 | Improved |
| remove self-refinement data (ablation) | ToolACE-R · BFCL Overall | accuracy: 85.04 -> 84.42 | Improved |
| add adaptive self-refinement at inference (ablation) | ToolACE-R · BFCL Overall | accuracy: 86.33 -> 86.83 | Improved |
| add adaptive self-refinement at inference | ToolACE-R · ACEBench | accuracy: 83.88 -> 84.00 | No clear effect |
| add adaptive self-refinement at inference | ToolACE-R · API-Bank | accuracy: 75.89 -> 81.22 | Improved |
| include extremely easy samples (alpha=-1) (ablation) | ToolACE-R · BFCL Overall | accuracy: 86.33 -> 86.00 | Hurt |
| keep only difficult samples (alpha=0.1, beta=2) (ablation) | ToolACE-R · BFCL Overall | accuracy: 86.33 -> 85.31 | Hurt |
| use k=8 attempts instead of k=1 for difficulty (ablation) | ToolACE-R · BFCL Overall | accuracy: 85.96 -> 86.33 | Improved |
| select data with the model being trained instead of Qwen2.5-3B (ablation) | Llama3.1-8B-Inst · BFCL Overall | accuracy: 85.50 -> 86.33 | Improved |
| use adaptive stopping instead of vanilla self-refine (ablation) | ToolACE-R · BFCL | accuracy | Improved |
| apply ToolACE-R training to Mistral-7B-Inst-v0.2 vs original-data SFT | Mistral-7B-Instruct-v0.2 · BFCL Overall | accuracy: 77.12 -> 84.00 | Improved |
| apply ToolACE-R training to Qwen2.5-7B-Inst vs original-data SFT | Qwen2.5-7B-Instruct · BFCL Overall | accuracy: 83.34 -> 85.21 | Improved |
Closest papers by meaning
- Learning Evolving Tools for Large Language Models (ICLR 2025 · similarity 0.92)
- CITI: Enhancing Tool Utilizing Ability in Large Language Models Without Sacrificing General Performance (AAAI 2025 · similarity 0.90)
- Joint Optimization of Tool Creation and Use for Large Language Model Agents (NeurIPS 2026 · similarity 0.89)
- ToolACE: Winning the Points of LLM Function Calling (ICLR 2025 · similarity 0.89)
- ToolRL: Reward is All Tool Learning Needs (NeurIPS 2025 · similarity 0.89)
- Unleashing the Potential of Large Language Models as Prompt Optimizers: Analogical Analysis with Gradient-based Model Optimizers (AAAI 2025 · similarity 0.88)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
