Detect cross-claim contradictions in knowledge graphs using DeBERTa NLI.
npx skills add https://github.com/mcleanT/AutoReview --skill nli-score
Please help me install this Agent Skill. Skill: nli-score Source: https://github.com/mcleanT/AutoReview/tree/main/.claude/skills/nli-score Command: npx skills add https://github.com/mcleanT/AutoReview --skill nli-score
This Skill automates detection of contradictions between claims in a literature-derived knowledge graph, reducing manual evidence triage and producing evidence-weighted confidence metrics for each claim.
Run the nli-score pipeline on your knowledge graph file to score claim contradictions and generate an interactive HTML and JSON report.
High-intent search queries and answers about installing and using this skill.
To detect contradictions in a knowledge graph, run a DeBERTa cross-encoder NLI pipeline over claim pairs to apply Beta-Binomial posterior updates and score cross-claim contradictions.
NLI classification identifies contradictions by running a DeBERTa cross-encoder over literature-extracted claim pairs and applying Beta-Binomial updates to per-edge posteriors for evidence-weighted confidence scoring.
Yes, contradiction detection supports MPS, CUDA, and CPU devices for batch inference, allowing you to scale NLI classification across large knowledge graphs built from extracted paper claims.
To filter false positives during contradiction detection, the pipeline skips parallel assertions and applies deterministic predicate-opposition checks to avoid spurious contradictions before running NLI classification.
The best way to visualize contradictions is by generating interactive HTML and JSON reports, which provide exploration interfaces and evidence-level diagnostics for analyzing cross-claim contradictions.
The contradiction detection pipeline skips parallel assertions as a false-positive pre-filtering step, applying deterministic predicate-opposition checks to ensure only genuine cross-claim contradictions are scored.
BoomberAspDraft and refine research papers with structured evidence and literature management.
wentoraiConduct systematic research cycles with gap identification and contradiction handling.
synaptiaiTransform research pillar evidence into structured SYN-<pillar>.md synthesis reports.
hellonishPair claims from distinct sources to detect and characterize cross-source contradictions.
