ImplicitMemBench - Classical Conditioning: leaderboard
Metric: First-action accuracy (%) on the 100 classical-conditioning items: after stimulus and consequence pairings and unrelated turns, the first action on the conditioned stimulus is judged by GPT-4o-mini for learned protective behavior (learning, interference and test phases, first-attempt scoring, mean of three runs at temperature 0); higher is better. Source: arxiv.org. Saturation forecast: Around 2029. 17 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | DeepSeek R1 | 69.67 |
| 2 | Qwen 3 32B | 67 |
| 3 | GPT-5 | 64 |
| 4 | Qwen 3 8B | 64 |
| 5 | O4 Mini (High) | 60 |
| 6 | O3 | 57.67 |
| 7 | GLM-4.5 | 53.33 |
| 8 | Claude Sonnet 4 | 51.67 |
| 9 | Gemini 2.5 Flash | 49 |
| 10 | Gemini 2.5 Pro | 47.33 |
| 11 | Llama 3.3 70B Instruct | 47.33 |
| 12 | Qwen 2.5 72B Instruct | 47 |
| 13 | GPT-4o Mini | 44 |
| 14 | GPT-4o | 43.67 |
| 15 | Claude Opus 4.1 | 41.67 |
Interactive version: theaggregate.ai/benchmark?slug=implicitmembench-classical-conditioning · How It Works · Data refreshed daily, snapshot 2026-10-07.