NoisyBench - No Distractor: leaderboard
Clean control of NoisyBench (Lost in the Noise; Lee et al., KAIST and LG AI Research, 2026): 2,766 items from 11 datasets in RAG (SealQA, MultihopRAG, Musique), reasoning (BBEH-Mini, AIME25, GPQA-Diamond), alignment (Self-Awareness, Survival-Instinct, BBQ) and tool use (TauBench Retail and Airline), with no distractor in the context. Each dataset keeps its own metric (pass@k, pass^k for TauBench) and the score is their harmonic mean in percent; higher is better.
Metric: Harmonic Mean Score (%; 11 datasets' pass@k, pass^k for TauBench; no distractor). Source: arxiv.org. Saturation forecast: Estimated already saturated. 7 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 2.5 Pro | 77.8 | #262 |
| 2 | DeepSeek R1 0528 | 72.4 | #425 |
| 3 | GPT-OSS-120B (High) | 72 | #699 (GPT-OSS-120B) |
| 4 | Gemini 2.5 Flash | 70.6 | #452 |
| 5 | Qwen 3 30B A3B 2507 (Thinking) | 58.8 | #759 |
| 6 | Qwen 3 4B 2507 (Thinking) | 58.4 | #1388 |
| 7 | DeepSeek R1 Distill Llama 8B | 32.4 | #6773 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=noisybench-no-distractor · How It Works · Data refreshed daily, snapshot 2026-11-12.