RobustBench-TC - Duplicate Tool Without Description: 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 distractor tool that shares the correct tool name but has no description or parameters; higher is better.

Metric: Tool-call accuracy (%) under the Dup-NoDesc perturbation (action component: inject distractor sharing GT name; no desc, no 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-32B65.4
2Qwen 3 32BNon-reasoning59.4#972
3LoopTool-32B59.4
4Qwen 3 14BNon-reasoning58.6#1269
5LoopTool-8B52.6#1473
6O4 Mini51.9#368
7Qwen 3.5 9BNon-reasoning51.9#828
8MUA-RL-8B50.4#1640
9MUA-RL-14B50.4#1658
10ToolRL-Qwen2.5-1.5B48.1#2299
11Qwen 3 8BNon-reasoning47.4#1873
12TL-CodeLLaMA-245.1
13ToolRL-Qwen2.5-3B45.1
14DeepSeek R1 Distill Qwen 14B41.4#2771
15Qwen 2.5 1.5B Instruct40.6#7470

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-without-description · How It Works · Data refreshed daily, snapshot 2026-11-21.