PersonalHomeBench - Counterfactual Reasoning (Agentic): leaderboard
Metric: Accuracy (%) on 2,000 counterfactual questions, macro-averaged over the variants that change and that keep the ground-truth answer (exact match, else a Gemini 2.5 Pro semantic-equivalence judge), over the text-based split of PersonalHomeBench (1,100 generated, manually reviewed households with persona-based occupants and appliances), temperature 0, in the agentic setting (the model must gather household information and act through the benchmark's PersonalHomeTools toolbox, up to 15 turns); each model at its stated reasoning mode; higher is better. Source: arxiv.org. Saturation forecast: Around January 2027. 11 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | GPT-OSS-20B | 63.65 |
| 2 | GPT-OSS-20B (High) | 63.25 |
| 3 | GPT-OSS-20B (Low) | 63.1 |
| 4 | GPT-4 | 62.9 |
| 5 | GPT-4o | 60.94 |
| 6 | nemotron-3-nano-30B-a3B | 57.95 |
| 7 | Qwen 3 4B (Non-reasoning) | 45.45 |
| 8 | Qwen 3 4B (Reasoning) | 36.35 |
Interactive version: theaggregate.ai/benchmark?slug=personalhomebench-counterfactual-reasoning-agentic · How It Works · Data refreshed daily, snapshot 2026-10-07.