response-level-hallucination-detection-against-reference-documents-having-model-first CLM-0218For response-level hallucination detection against reference documents, having the model first convert the reference into a relational SQL database and then check response claims by executing SELECT queries raises F1 over zero-shot direct prediction mainly by cutting false positives, and the gain shrinks or vanishes on backbones with weak text-to-SQL ability.
Capability: Stating false facts confidently
Observed on
RAGTruth (QA, data-to-text, summarization) and DiaHalu task-oriented dialogue; Gemini 2.5 Flash plus Qwen3, GPT-OSS, LLaMA 3.3 70B; unsupervised, no fine-tuning; five prompting ste.
Sources
- supportsLeveraging Low-Level Symbolic Competences for Unsupervised Grounding in Hallucination DetectionMacro F1 on RAGTruth rose from 61.6 (direct prediction) to 71.3 with the SQL pipeline on Gemini 2.5 Flash, with precision 48.2 to 61.7 and recall falling 95.4 to 88.9; atomic-fact and JSON key-value multi-step baselines landed between (65.7, 64.9), so part of the gain is decomposition rather than SQL specifically. GPT-OSS, lowest Spider EM, showed no gain (68.7 to 68.6). Single paper, one main backbone for the headline result, no fine-tuned control on DiaHalu.
Disagreeing is the most useful thing you can do here. Nothing on this page waits for a reviewer: the counts above move when evidence lands, and every source stays attached with who brought it.
Related claims
- For response-level hallucination detection in retrieval-augmented settings, having the model first convert the reference documents into a relational SQL database and then check the response against SELECT query results raises precision and macro F1 over zero-shot direct prediction, over atomic-claim decomposition, and over flat JSON key-value extraction, without any task-specific fine-tuning.Stating false facts confidently · reviewed by AI
- For instruction tuning an 8B open model on text-to-SQL, a curated set of under 900 examples selected to cover every table in the target schemas and score well on question-SQL alignment reaches higher execution accuracy than pipelines trained on tens of thousands to millions of samples, and dropping the schema-coverage term from the selection objective costs more accuracy than dropping any single quality term.Writing a reusable locator: regex, SQL, XPath · reviewed by AI
- On natural-language-to-SQL synthesis over multi-table databases, a model's accuracy overstates its ability whenever the same database or the same query appears in both training and test: splitting by database rather than by example cut the best system's exact-match accuracy from 33.0% to 12.4%, and accuracy falls further as the target database has more foreign keys.Writing a reusable locator: regex, SQL, XPath · reviewed by AI
- In synthetic hallucination benchmarks where faithful items are human-written and hallucinated items are LLM-rewrites, a detector's binary score partly measures human-vs-machine style rather than factuality: feeding faithful LLM rewrites of the same truthful articles makes some fine-tuned encoders flag nearly all of them as hallucinated.Whether the measurement made the finding · reviewed by AI
- For training-free inference-time hallucination mitigation in 7B vision- language models, reported gains on object-hallucination benchmarks are largely inseparable from reduced informativeness: hallucination rate and object recall/coverage move together, so a lower score can mean the model mentioned fewer visual entities rather than grounded better.Whether the measurement made the finding · reviewed by AI
Notes
Drafted from the paper by a model and filed unreviewed. Visible here so it can be read, not because anyone has vouched for it: it does not move any technique's standing and does not count toward the internal scorecard. Drafted confidence: medium. Falsifier as drafted: A run where SQL-grounded checking gives no precision gain over direct prediction on the same backbones and datasets, or where a backbone with low text-to-SQL exact match gains as much as a high-scoring one. Proposed technique, not catalogued: SQL-grounded hallucination checking.