IBA-Bench - Leisure Activities: leaderboard

Metric: Checkpoint satisfaction (%; share of hidden preference checkpoints satisfied by the final action, leisure activities domain; mean of three LLM judges: DeepSeek-V3.2, Qwen3-4B-Instruct and Qwen3-30B-A3B-Instruct; retrieval-augmented baseline with bge-m3 top-3 chunks). Source: arxiv.org. Saturation forecast: Around 2029. 10 models tracked.

Top models

#ModelScore
1GPT-5.156.8
2Qwen 3 4B Instruct52.6
3GPT-5 Mini49.8
4DeepSeek V3.248.3
5QwQ-32B47.3
6DeepSeek V344.7
7GPT-4o44.1
8Qwen 2.5 7B Instruct40.1
9GPT-4o Mini39.9
10GLM-4 9B Chat36.7

Interactive version: theaggregate.ai/benchmark?slug=iba-bench-leisure-activities · How It Works · Data refreshed daily, snapshot 2026-09-26.