RobustBench-TC - Duplicate Tool With Wrong Parameters: leaderboard

RobustBench-TC (2026) tests tool-calling models under deployment noise grounded in real tool-calling failures, using 199 single-turn samples drawn from BFCL, API-Bank, RoTBench, ToolAlpaca and ToolEyes and scored by deterministic rule-based parsers. This board is tool-call accuracy in percent under an injected same-name distractor tool with no description and wrong parameters; higher is better.

Metric: Tool-call accuracy (%) under the Dup-WrongP perturbation (action component: same name; no description, wrong params); RobustBench-TC samples drawn from BFCL V3 single-turn, API-Bank, RoTBench, ToolAlpaca and ToolEyes, each scored by its source benchmark's strict scorer after a format-tolerant tool-call parse, temperature 0 (DeepSeek-R1-Distill at 0.6), Qwen3-family models with thinking disabled; higher is better. Source: arxiv.org. Saturation forecast: Around February 2027. 18 models tracked. Updated 7 Oct 2026.

Top models

#ModelSettingScoreOverall rank
1MUA-RL-32B73.8
2LoopTool-32B69.8
3Qwen 3 32BNon-reasoning64.3#972
4Qwen 3 14BNon-reasoning57.1#1269
5ToolRL-Qwen2.5-1.5B56.3#2299
6O4 Mini54.8#368
7Qwen 3.5 9BNon-reasoning53.2#828
8LoopTool-8B53.2#1473
9ToolRL-Qwen2.5-3B52.4
10MUA-RL-14B51.6#1658
11MUA-RL-8B50#1640
12TL-CodeLLaMA-250
13Qwen 3 8BNon-reasoning48.4#1873
14ToolRL-Llama3.2-3B43.7
15DeepSeek R1 Distill Qwen 14B42.9#2771

No result here: #1 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.

Interactive version: theaggregate.ai/benchmark?slug=robustbench-tc-duplicate-tool-with-wrong-parameters · How It Works · Data refreshed daily, snapshot 2026-11-21.