model-asked-audit-many-documents-planted-errors-single-large-batch CLM-0441When a model is asked to audit many documents for planted errors in a single large-batch prompt, recall of known contaminants collapses and the model reports confident fabricated findings — invented absurd contaminants that appear in no document — rather than signalling incomplete processing.
Capability: Stating false facts confidently
Observed on
Gemini 3.0 Pro auditing 60 academic papers with 180 known planted contaminants (typographical, semantic reversal, absurd insertion), compared across single-document, small-batch an.
Sources
- Validated by: Not measured; the paper only argues for it from the observed batch-size degradation.. Measured on one model and one corpus: 180-contaminant answer key over 60 documents, 50% recovery single-document, 60% small batch, 2.8% large batch, with named fabricated items. No comparison model, small sample, and the single-vs-small-batch difference is unresolved by the authors' own admission.
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
- When training an 8B multimodal model with RL to find scientific errors in whole papers without being told where to look, mixing in matched easier variants of the same instances — where the error is asserted and the evidence supplied — yields better performance on the evidence-absent task itself than training only on that task at matched exposure.Spotting an error in a given artifact · reviewed by AI
- For persistent-memory agents, any memory-write rule that decides using only recency and provenance is stuck on a single tradeoff — accepting more genuine preference updates means admitting more poisoned ones — because an adversary who can launder a claim through the user's own channel matches the statistics a genuine revision produces; conditioning the write decision on the inferred authenticity of the claim given the interaction history moves both axes at once.Remembering across sessions · reviewed by AI
- Auditing AI-generated code with static smell detection at code, structure and architecture level, from single-file tasks up to agent-built systems, more capable models produced more bloated and more coupled code, code volume was a near-perfect predictor of structural decay, and neither functional correctness nor more detailed prompting reduced it, so passing tests does not indicate maintainable design.Making design decisions that hold up as a system grows · reviewed by AI
- 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
- In LLM fact-checking over supplied evidence, high veracity accuracy does not imply the verdict depends on that evidence: when the model's own cited sentences are iteratively removed, off-the-shelf models (GPT-4o-mini, Gemini-2.5-Flash-Lite, Qwen-2.5-32B, Llama-3.1-8B) keep their SUPPORT/REFUTE labels at a non-zero plateau instead of abstaining, and training with paired supervision that requires NOT_ENOUGH_INFO on the evidence-ablated input removes most of that persistence without losing accuracy.Checking claims against evidence · 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 model asked to audit a large batch of documents recovers planted contaminants at a rate comparable to single-document auditing, or reports incomplete coverage instead of emitting contaminants absent from every document.