AI Paper Atlas

Research map / Large language models / Agents, code and math

Gecko: A Simulation Environment with Stateful Feedback for Refining Agent Tool Calls

Zeyu Zhang, Guohao Li, Zhenchang Xing, Alexandros Apostolopoulos, Yu Lin Lee, Liang Zheng

ICML 2026 · Large language models › Agents, code and math · approach: Agent instruction-following benchmarks · 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 GATS (simulated refinement in Gecko) to the planning LLMGPT-4o · BFCLv3 overalloverall accuracy (%): 76.93 -> 84.62Improved
add GATS to the planning LLMGPT-5 · BFCLv3 overalloverall accuracy (%): 61.94 -> 66.08Improved
add GATS to the planning LLMGemini-3.0-pro · BFCLv3 overalloverall accuracy (%): 79.97 -> 85.19Improved
add GATS to the planning LLMGPT-4o · tau2-bench overallaverage pass@1 (%): 54.2 -> 60.7Improved
add GATS to the planning LLMGPT-5.5 · tau2-bench overallaverage pass@1 (%): 85.5 -> 87.9Improved
add GATS to the planning LLMDeepseek-V3 · BFCLv3 overalloverall accuracy (%): 70.40 -> 72.90Improved
add GATS to GPT-4o for multi-turn baseGPT-4o · BFCLv3 multi-turn baseaccuracy (%): 61.0 -> 72.0Improved
add GATS rather than Reflexion, Self-Refine, Best-of-N or Majority VotingGPT-5-mini · tau2-airlinepass@1 (%): 60.0 -> 65.0Improved
collapse Gecko into a single prompt (Merge-to-one) (ablation)GPT-4o · BFCLv3 multi-turn baseaccuracy (%): 61.0 -> 58.0Hurt
add task feedback judge LLM (ablation)Gecko · BFCLv3 multi-turn baseaccuracy (%): 61.5 -> 72.0Improved
add task state estimation (ablation)Gecko · BFCLv3 multi-turn baseaccuracy (%): 68.0 -> 72.0Improved
add argument validation (ablation)Gecko · BFCLv3 multi-turn baseaccuracy (%): 70.5 -> 72.0Improved

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