Themis-CodeRewardBench - Execution Efficiency: leaderboard
Metric: Preference accuracy (%; share of the benchmark's preference pairs in which the reward model scores the preferred code response above the rejected one, pointwise and reference-free; Execution Efficiency criterion, 1,309 pairs: runtime-improving commits, Pie4Perf, ECCO and EvalPerf). Source: arxiv.org. Saturation forecast: Estimated already saturated. 51 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | LDL-Reward-Gemma-2-27B-v0.1 | 65.47 |
| 2 | Llama-3.1-Tulu-3-70B-SFT-RM-RB2 | 65.16 |
| 3 | Skywork-Reward-V2-Qwen3-8B | 64.63 |
| 4 | Llama 3.1 70B Instruct RM RB2 | 64.02 |
| 5 | INF-ORM-Llama3.1-70B | 62.03 |
| 6 | ArmoRM-Llama3-8B-v0.1 | 61.96 |
| 7 | Llama 3.1 8B Base RM RB2 | 61.42 |
| 8 | URM-LLaMa-3.1-8B | 60.73 |
| 9 | InternLM2-20B-Reward | 60.2 |
| 10 | QRM-Llama3.1-8B-v2 | 59.74 |
| 11 | FsfairX-LLaMA3-RM-v0.1 | 59.74 |
| 12 | Llama-3-OffsetBias-RM-8B | 59.05 |
| 13 | GRM-llama3-8B-sftreg | 58.75 |
| 14 | InternLM2-7B-Reward | 57.37 |
| 15 | AceCodeRM-7B | 55.54 |
Interactive version: theaggregate.ai/benchmark?slug=themis-coderewardbench-execution-efficiency · How It Works · Data refreshed daily, snapshot 2026-09-26.