AsymLoop 35B Tokens - Zero-shot - ARC-Easy: leaderboard
AsymLoop (2026):350M-scale research checkpoints trained on35Btokens, distinct from7B-token main experiment; matched effective depth, LM Evaluation Harness Zero-shot. Accuracy (%).
Metric: Accuracy (%). Source: arxiv.org. Saturation forecast: Rough model projection: around 2026. 3 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | AsymLoop Study 350M AsymLoop K3 Base 35B tokens | 68.35 | |
| 2 | AsymLoop Study 350M Dense Base 35B tokens | 66.33 | |
| 3 | AsymLoop Study 350M Full Loop K2 Base 35B tokens | 66.2 |
Interactive version: theaggregate.ai/benchmark?slug=asymloop-35b-tokens-zero-shot-arc-easy · How It Works · Data refreshed daily, snapshot 2026-10-22.