Plan-RewardBench - Tool Irrelevance: leaderboard
Plan-RewardBench (2026) tests reward models and LLM judges on preferring the better of two tool-using agent trajectories, with preference labels from a judge panel audited by humans. This board is pairwise accuracy in percent on the pairs where the tools are irrelevant or unavailable; higher is better.
Metric: Pairwise accuracy (%) on the 275 tool-irrelevance or tool-unavailable pairs of the Plan-RewardBench chosen versus rejected agent trajectories (shared tool registry, multi-turn context, executed tool feedback), preference labels from a Gemini 3 Pro and GPT-5.1 judge panel with human audit; discriminative reward models score each trajectory, generative reward models and LLM judges choose between the pair; higher is better. Source: arxiv.org. Saturation forecast: Around June 2027. 21 models tracked. Updated 7 Oct 2026.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3 Flash (Preview) | 75.55 | #152 |
| 2 | GPT-5 | 69.73 | #178 |
| 3 | RM-R1-DS-Distill-Qwen-32B | 69.1 | |
| 4 | Qwen 3 235B A22B 2507 (Thinking) | 67.36 | #466 |
| 5 | INF-ORM-Llama3.1-70B | 66.91 | |
| 6 | Skywork-Reward-V2-Qwen3-8B | 65.45 | |
| 7 | Skywork-Reward-V2-Llama-3.1-8B | 64.91 | |
| 8 | M-Prometheus-14B | 63.89 | #1255 |
| 9 | Qwen Plus | 63.73 | #820 |
| 10 | InternLM2-7B-Reward | 62.91 | |
| 11 | RRM-32B | 61.15 | #1115 |
| 12 | Qwen 3 4B 2507 Instruct | 60.27 | #2221 |
| 13 | DeepSeek V3.2 Exp | 60 | #439 |
| 14 | Qwen Max | 60 | #1092 |
| 15 | Qwen 3 235B A22B 2507 Instruct | 59.82 | #616 |
No result here: #1 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.
Interactive version: theaggregate.ai/benchmark?slug=plan-rewardbench-tool-irrelevance · How It Works · Data refreshed daily, snapshot 2026-11-21.