HELM SeaHELM - LINDSEA Scalar Implicatures (id): leaderboard
HELM SeaHELM LINDSEA Indonesian scalar-implicature task (exact match).
Metric: EM. Source: crfm.stanford.edu. Saturation forecast: Estimated already saturated. 21 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Llama 3.3 70B Instruct | 92.31 | #534 |
| 2 | Llama 3.1 70B Instruct | 91.35 | #561 |
| 3 | DeepSeek V3 | 87.5 | #321 |
| 4 | Claude 3.7 Sonnet (20250219) | 87.5 | #201 |
| 5 | Qwen 2.5 72B Instruct | 86.54 | #454 |
| 6 | Llama 4 Scout Instruct | 86.54 | #539 |
| 7 | Gemini 2.0 Flash Lite | 86.54 | #450 |
| 8 | Qwen 2 72B Instruct | 85.58 | #576 |
| 9 | Gemini 2.0 Flash | 80.77 | #342 |
| 10 | Llama 3.2 90B Vision Instruct | 74.04 | #622 |
| 11 | GPT-4o (2024-11-20) | 72.12 | #381 |
| 12 | Llama 4 Maverick Instruct FP8 | 48.08 | #414 |
| 13 | Llama 3.1 405B Instruct | 42.31 | #457 |
| 14 | GPT-4o Mini (2024-07-18) | 25.96 | #558 |
| 15 | Gemma 2 27B (IT) | 16.35 | #645 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=helm-seahelm-lindsea-scalar-implicatures-id · How It Works · Data refreshed daily, snapshot 2026-10-20.