KARMA: Knowledge graph-based Automated Reasoning Materialization and Alignment
Quick Answer
KARMA addresses the Resolution Mismatch Problem in template-based contrastive synthesis by utilizing schema-constrained paths in knowledge graphs for slot-aligned candidates.
Quick Take
KARMA addresses the Resolution Mismatch Problem in template-based contrastive synthesis by utilizing schema-constrained paths in knowledge graphs for slot-aligned candidates. It employs Slot-Parallel Alignment (SPA) to enhance preference supervision, outperforming base LLMs and same-data SFT baselines across biomedical, computer science, and chemistry benchmarks.
Key Points
- KARMA formalizes the Resolution Mismatch Problem in contrastive synthesis.
- It generates slot-aligned candidates using knowledge graphs.
- Slot-Parallel Alignment (SPA) improves preference supervision.
- KARMA outperforms base LLMs and SFT baselines in multiple benchmarks.
- Effective across biomedical, computer science, and chemistry domains.
Paper Resources
📖 Reader Mode
~2 min readAbstract: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.
| Comments: | First version, 20 pages |
| Subjects: | Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Machine Learning (cs.LG) |
| Cite as: | arXiv:2607.03166 [cs.CL] |
| (or arXiv:2607.03166v1 [cs.CL] for this version) | |
| https://doi.org/10.48550/arXiv.2607.03166 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Jinkyeong Choi [view email]
[v1]
Fri, 3 Jul 2026 10:06:40 UTC (654 KB)
— Originally published at arxiv.org
Want this in your inbox every morning?
Daily brief at your local 8am — bilingual EN/中文, free.
More from arXiv cs.CL
See more →Quantifying Prior Dominance in Systems
The study introduces the Normalized Context Utilization (NCU) metric to evaluate Retrieval-Augmented Generation (RAG) systems, revealing that Small Language Models (SLMs) outperform larger models in factual extraction. The findings indicate that traditional scaling laws yield diminishing returns, with a commercial API frequently failing against adversarial evidence due to systemic confidence collapse.