paper_type.json (345B)
1 { 2 "paper_type": "empirical", 3 "reason": "Runs experiments evaluating LLM performance on hierarchical legal reasoning, reports quantitative results (100% on surface-level vs 11-34% on integrated analysis, 2.6x RL improvement), with primary contribution being empirical findings about the thinking-longer paradox rather than a new benchmark." 4 }