TBNAI
KDASO: Knowledge-Data Dual-Driven Automated Skill Optimization for Liability Adjudication Tasks
Publication: cs.AI updates on arXiv.org
Author: Zi-Jian Cheng, Weiming Wu, Jie Meng, Zhen Peng, Shan Huang, Qun Li, Guobing Wu, Lan-Zhe Guo
Source published: 06/11/2026, 05:00:00 UTC
Briefing created: 06/11/2026, 06:04:07 UTC
Based on a feed excerpt; full article content has not been acquired. Source evidence retrieved 06/11/2026, 05:03:24 UTC.
Briefings
KDASO is a new system designed to continuously improve legal liability adjudication skills by analyzing recent case data and LLM reasoning trajectories. Experiments show KDASO significantly outperforms general-purpose and domain-specific baselines, achieving an 8.96% improvement when transferring skills to concept-similar scenarios. The system uses neighborhood-conditioned induction from adjudication data and execution-aligned refinement based on LLM trajectory analysis.
Key points
- KDASO uses neighborhood-conditioned induction to continuously evolve a skill library from recent adjudication data.
- KDASO achieves 8.96% improvement in liability adjudication tasks when transferring skills to concept-similar scenarios.
Context
Liability adjudication tasks require analyzing event descriptions against domain-specific skills to determine liability. Existing approaches face limitations including stale skill libraries, semantic execution gaps, and context budget overflows. This finding matters as it offers a scalable solution for dynamic legal reasoning tasks.
Limitations
Evidence covers three specific domains: ride-hailing adjudication, legal liability determination, and logical reasoning judgment. Results are from experiments demonstrating performance gains, not a full-scale deployment. The 8.96% improvement is a transfer metric to concept-similar scenarios, not a direct comparison to all current legal systems. No specific city, region, or country is identified in the source evidence.
Generated by Bittrees-hosted qwen35-2b sha256:aaf42c8b7c3c on 06/11/2026, 06:04:07 UTC. Based on publisher-provided feed evidence; may not cover the full original article. Check the source for details.
View archived briefing on IPFS
bafybeidbe35yggi7qlih4eicwokybt24rzxyjnh4fku2bwfckiugf5h3eu