BFI-Adapt: leaderboard
BFI-Adapt (Northeastern University and Singapore Management University, 2026) scores whether personality-conditioned LLM agents change their Big Five inventory answers after simulated life events in the directions human longitudinal studies report. The composite, in percent, combines item-level rating reliability, within-pair directional consistency and agreement with the human prior over 27 event-trait pairs; higher is better.
Metric: BFI-Adapt composite (%; mean over the 27 event-trait pairs with a definite human change direction of max(0, weighted kappa) x (2 DCR - 1) x an indicator that the dominant item-change direction matches the human prior; 100 personas simulate 11 life events and answer the BFI-44 before and after; non-reasoning mode). Source: arxiv.org. Saturation forecast: Around July 2028. 14 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3 Flash (Minimal) | 34.8 | #159 (Gemini 3 Flash) |
| 2 | GLM-4.6 (Non-reasoning) | 32.2 | #485 (GLM-4.6) |
| 3 | Qwen 3 235B A22B (Non-reasoning) | 28.7 | #622 (Qwen 3 235B A22B) |
| 4 | Claude Haiku 4.5 | 22.5 | #568 |
| 5 | Kimi K2 0905 | 22.4 | #555 |
| 6 | Claude Sonnet 4.6 | 21.2 | #158 |
| 7 | Qwen 3.5 9B (Non-reasoning) | 19.4 | #803 (Qwen 3.5 9B) |
| 8 | GPT-4.1 Mini | 19.2 | #775 |
| 9 | GPT-5.3 | 17.8 | #194 |
| 10 | DeepSeek V4 Pro (Non-reasoning) | 17.6 | #181 (DeepSeek V4 Pro) |
| 11 | MiMo-V2.5-Pro (Non-reasoning) | 7.1 | #274 (MiMo-V2.5-Pro) |
| 12 | Qwen 2.5 14B Instruct | 4.9 | #1745 |
| 13 | InternLM3-8B-Instruct | 4.4 | #2586 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=bfi-adapt · How It Works · Data refreshed daily, snapshot 2026-11-12.