SWE-QA (Multi-Hop Code) - Interacting Entities: leaderboard
Metric: Accuracy (%) on the 4,488 Interacting-Entity questions, which follow two entities interacting across three chunks, of SWE-QA, multi-hop multiple-choice questions (four options) about the source of 12 SWE-bench Python repositories, generated with Llama-3.2-3B-Instruct; oracle setting (only the relevant code chunks), zero-shot, labels after the 66-item consensus correction; chance 25; higher is better. Source: arxiv.org. Saturation forecast: Around January 2027. 15 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Llama 3.3 70B Instruct | 69.98 |
| 2 | Gemma 3 4B (IT) | 68.6 |
| 3 | Qwen 3 4B Instruct | 65.41 |
| 4 | GPT-OSS-20B | 53.83 |
| 5 | GPT-OSS-20B (Low) | 53.43 |
| 6 | GPT-OSS-20B (High) | 53.16 |
| 7 | Phi-4-mini (Reasoning) | 50.35 |
| 8 | Phi-4 Mini Instruct | 47.9 |
| 9 | Qwen 3 1.7B | 43.18 |
| 10 | DeepSeek R1 Distill Qwen 1.5B | 39.14 |
| 11 | SmolLM2-1.7B-Instruct | 29.94 |
| 12 | SmolLM2-360M-Instruct | 20.61 |
Interactive version: theaggregate.ai/benchmark?slug=swe-qa-multi-hop-code-interacting-entities · How It Works · Data refreshed daily, snapshot 2026-10-07.