NLU Corruption - New Items MNLI Style: leaderboard
This study tests natural-language inference with clean and corrupted inputs. Raw accuracy, answer-bias-centred accuracy and prompting conditions remain separate. The public ANLI-trained RoBERTa checkpoint is distinct from generic RoBERTa and the study’s locally fine-tuned encoders. Above-chance retention is not relabelled as accuracy. This board reports clean answer accuracy, centered where probabilities exist (%). Higher is better.
Metric: Clean answer accuracy, centered where probabilities exist (%). Source: arxiv.org. Saturation forecast: Estimated already saturated. 8 models tracked. Updated 14 Nov 2026.
Top models
| # | Model | Setting | Score | Overall rank |
|---|---|---|---|---|
| 1 | Qwen 3.8 27B | 100 | #209 | |
| 2 | DeepSeek V4 Pro | 99.4 | #191 | |
| 3 | OLMo 3 32B Base | 99.4 | ||
| 4 | Muse Glimmer 30B | 98.8 | #391 | |
| 5 | GPT-4o | 98.2 | #744 | |
| 6 | Mistral Large 3 | 98.2 | #827 | |
| 7 | OLMo 3 7B Base | 98.2 | ||
| 8 | Grok 4.7 | Reasoning | 98.2 | #47 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=nlu-corruption-new-items-mnli-style · How It Works · Data refreshed daily, snapshot 2026-11-16.