internlm2-chat-20B-ExPO: benchmark results
Community ExPO variant of InternLM2-chat-20B (2024) extrapolating SFT and RLHF weights for stronger alignment without extra training. Provider: chujiezheng. Released 2024-05-02. Access: Open.
Aggregate Rating 1502 ± 38, rank #2317 of 10916 rated models, from 9 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Open Chinese LLM - ARC Challenge | 59.47 | Accuracy (%) | 84.8 |
| Open Chinese LLM - HellaSwag | 65.95 | Accuracy (%) | 80.2 |
| Open Chinese LLM - WinoGrande | 66.22 | Accuracy (%) | 77.1 |
| Open Chinese LLM Leaderboard | 61.26 | Official Nonzero-Task Mean (%) | 71.6 |
| Open Chinese LLM - C-Eval Semantic | 81.42 | Accuracy (%) | 70.4 |
| Open Chinese LLM - CMMLU | 59.05 | Accuracy (%) | 67.2 |
| AlpacaEval 2.0 | 27.23 | LC Win Rate (%) | 64 |
| Open Chinese LLM - GSM8K | 45.64 | Accuracy (%) | 49.4 |
| Open Chinese LLM - TruthfulQA MC | 51.07 | Accuracy (%) | 25 |
Interactive version: theaggregate.ai/model?slug=internlm2-chat-20b-expo · How It Works · Data refreshed daily, snapshot 2026-11-08.