RobustBench-TC - Misleading Description (Neutral Suffix): 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 a misleading description on the correct tool and a neutral suffix on the distractor; higher is better.

Metric: Tool-call accuracy (%) under the MisDesc-N perturbation (reward component: MisDesc + neutral suffix on distractor (_1)); 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
1Qwen 3 32BNon-reasoning63.7#983
2O4 Mini54.9#372
3LoopTool-8B48#1497
4Qwen 3.5 9BNon-reasoning47.1#838
5DeepSeek R1 Distill Qwen 14B46.1#2739
6MUA-RL-8B45.1#1656
7Qwen 3 8BNon-reasoning44.1#1890
8MUA-RL-14B44.1#1674
9LoopTool-32B42.2
10MUA-RL-32B42.2
11Qwen 3 14BNon-reasoning40.2#1284
12ToolRL-Qwen2.5-1.5B40.2#2330
13ToolRL-Qwen2.5-3B39.2
14TL-CodeLLaMA-235.3
15Qwen 2.5 3B Instruct32.4#5410

No result here: #2 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.

Interactive version: theaggregate.ai/benchmark?slug=robustbench-tc-misleading-description-neutral-suffix · How It Works · Data refreshed daily, snapshot 2026-11-22.