Delulu - Exact Match: leaderboard
Metric: Exact match (%) of the completion with the gold completion, byte for byte, on all 1,951 Delulu fill-in-the-middle samples (seven languages, four injected hallucination types), greedy decoding, at most 256 tokens; base FIM models are truncated to the gold completion's line count; higher is better. Source: arxiv.org. Saturation forecast: Estimated already saturated. 11 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Qwen 2.5 Coder 32B Instruct | 54.1 |
| 2 | CodeLlama-13B-hf | 47.3 |
| 3 | CodeLlama-7B-hf | 46.8 |
| 4 | Qwen 2.5 Coder 14B Instruct | 46.2 |
| 5 | Qwen 2.5 Coder 3B Instruct | 44 |
| 6 | Qwen 2.5 Coder 7B Instruct | 40.2 |
| 7 | Qwen2.5-Coder-1.5B-Instruct | 37.8 |
| 8 | starcoder2-15B | 21.5 |
| 9 | starcoder2-7B | 9.7 |
Interactive version: theaggregate.ai/benchmark?slug=delulu-exact-match · How It Works · Data refreshed daily, snapshot 2026-10-07.