AI Paper Atlas

Research map / Reinforcement learning / RL from feedback

On Effectiveness and Efficiency of Agentic Tool-calling and RL Training

Tong Liu, Cheng Qian, Matej Cief, Yuan He, Daniele Dan, Nikolaos Aletras, Gabriella Kazai

ICML 2026 · Reinforcement learning › RL from feedback · approach: RL for LLM reasoning · 0 citations (OpenAlex)

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 changedWhereMeasuredEffect
add online pre-rollout filtering and max-variance rollout down-sampling to GRPOGRPO, Qwen2.5-3B-Instruct single-turn · BFCL single-turn tool-callingtraining speedup at comparable performance: 1.7x speedupImproved
add online pre-rollout filtering and max-variance rollout down-sampling to GRPOGRPO, Qwen3-4B multi-turn · BFCL multi-turn tool-callingtraining speedup at comparable performance: 2.6x speedupImproved
use native multi-turn template instead of context concatenation (ablation)Qwen3-8B, Qwen3-4B, Qwen2.5-7B-Instruct · BFCL multi-turnaccuracy: about 6-8% gainImproved
retain thinking history across turns (ablation)Qwen3 models · BFCL multi-turnaccuracy: about 3-5% for Qwen3-8B and Qwen3-4BImproved
add multi-turn instructions to the default BFCL system prompt (ablation)Qwen3-4B · BFCL multi-turnaccuracy: 22.7 -> 37.2Improved
train on multi-turn data instead of single-turn data (ablation)Qwen3-4B · BFCL multi-turnaccuracy: 22.7 -> 15.9Hurt
train on single-turn data (ablation)Qwen3-4B · BFCL multi-turnaccuracy: 22.7 -> 20.2No clear effect
evaluate with multiple random seeds instead of one run (ablation)BFCL evaluation of five models · BFCL multi-turnscore variance: up to ~3% deviation in multi-turnMixed
apply RL tool-calling trainingQwen3-4B · ACEBench English overalloverall accuracy: 65.4 -> 77.5Improved
apply RL tool-calling training on top of stronger system promptQwen3-4B · BFCL multi-turnaccuracy: 37.2 -> 39.4Improved
apply tool-calling RL trainingQwen3-4B · MMLU, HellaSwag, TruthfulQA, Winograndeaccuracy: MMLU 68.3 -> 68.3No clear effect

Closest papers by meaning

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.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy