KARMA: Knowledge graph-based Automated Reasoning Materialization and Alignment

Researchers propose KARMA, a method that uses knowledge graphs to generate slot-aligned contrastive candidates for preference optimization in large language models, addressing the Resolution Mismatch Problem. KARMA outperforms base LLMs and supervised fine-tuning baselines across biomedical, computer science, and chemistry benchmarks.

arXiv:2607.03166v1 Announce Type: new Abstract: Template-based contrastive synthesis is scalable, but its candidates often differ only in a few entity-slots while sequence-level optimization spreads supervision over mostly shared templates. We formalize this as the Resolution Mismatch Problem and propose KARMA, which enumerates schema-constrained paths over domain knowledge graphs and verbalizes them into slot-aligned contrastive candidates. Slot-Parallel Alignment SPA then applies a decoupled slot-level objective to route preference supervision to discriminative entity-slots, with slot-aware masked attention serving as an optional packed-evaluation implementation. Across biomedical, computer-science, and chemistry benchmarks, KARMA outperforms base LLM and same-data SFT baselines, and compares favorably with sequence and token-level preference methods.