RoundTripCodeEval - Run-Length Encoding: leaderboard
Metric: Mean of exact match, edit similarity and pass@5 (%) over the four round-trip tasks for run-length encoding (output prediction and input prediction, each directly and through the inverted function), 250 inputs, zero-shot with one worked example, five completions. Source: arxiv.org. Saturation forecast: Estimated already saturated. 15 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | QwQ-32B | 57.23 |
| 2 | Qwen 2.5 Coder 32B Instruct | 41.51 |
| 3 | DeepSeek R1 Distill Qwen 32B | 36.37 |
| 4 | Codestral-22B-v0.1 | 30.68 |
| 5 | DeepSeek R1 Distill Qwen 14B | 26.97 |
| 6 | Qwen 2.5 7B Instruct | 17.39 |
| 7 | Phi-3.5-mini-instruct | 13.67 |
| 8 | Phi-3 Mini 128K Instruct | 12.01 |
| 9 | Yi-Coder-9B-Chat | 11.85 |
| 10 | Mistral 7B Instruct (v0.3) | 11.52 |
| 11 | Llama 3.1 8B Instruct | 9.36 |
| 12 | DeepSeek R1 Distill Qwen 1.5B | 4.3 |
| 13 | Llama 3.2 1B Instruct | 0.15 |
Interactive version: theaggregate.ai/benchmark?slug=roundtripcodeeval-run-length-encoding · How It Works · Data refreshed daily, snapshot 2026-09-26.