NoisyBench - Random Chat History: leaderboard
NoisyBench (Lost in the Noise; Lee et al., KAIST and LG AI Research, 2026) with an irrelevant multi-turn chat history sampled from WildChat placed before each of 2,766 items from 11 RAG, reasoning, alignment and tool-use datasets. 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; a random WildChat chat history added). Source: arxiv.org. Saturation forecast: Around February 2027. 7 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 2.5 Pro | 62.5 | #262 |
| 2 | DeepSeek R1 0528 | 59.4 | #425 |
| 3 | Gemini 2.5 Flash | 56.9 | #452 |
| 4 | GPT-OSS-120B (High) | 54.9 | #699 (GPT-OSS-120B) |
| 5 | Qwen 3 30B A3B 2507 (Thinking) | 49.3 | #759 |
| 6 | Qwen 3 4B 2507 (Thinking) | 46.5 | #1388 |
| 7 | DeepSeek R1 Distill Llama 8B | 23 | #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-random-chat-history · How It Works · Data refreshed daily, snapshot 2026-11-12.