Live intelligence

Latest AI updates

2 updates
Research

CLADE Shows RAG Models Guess From Coarse Evidence Instead of Abstaining

Researchers identify a failure mode called CLADE, where retrieval-augmented language models answer from passages that mention only a broader category rather than the specific answer required. On biomedical questions from PrimeKG, family-only passages reduced abstention by 13.8 points versus uninformative passages, and the resulting guesses were wrong about half the time. A counterfactual test replacing an exact disease name with its broader family lowered abstention by 17.4 points, while confidence filtering, larger models, and abstention instructions did not resolve the issue. Ontology-assisted recovery improved end-to-end performance from 10.0% to 26.8%, but identifying the correct family on the page remained a bottleneck.

retrieval-augmented generationabstentionbiomedical ontologies
Research

From Fluent Text to Ontology-Compatible Definitions: A Quality-Control Agentic Workflow

The paper addresses the gap between fluent LLM-generated biomedical ontology definitions and definitions that are compatible with ontology entities and logical constraints. It introduces Entity Error Rate and Logical Inconsistency Rate to measure entity grounding and logical consistency, complementing reference-based text metrics. An agentic quality-control workflow iteratively generates definitions, extracts axioms, checks entity existence and logical consistency, and uses diagnostics for repair. Experiments on Disease Ontology, Uberon, and Uberon-ConstraintStress show reduced Entity Error Rate while conventional text-generation performance remains strong; Logical Inconsistency Rate reveals failures that similarity metrics miss, though repair remains limited.

LLMsbiomedical ontologiesagentic workflows