# Gemma-3-12b fails to detect misinformation when poisoned claims dominate retrieved evidence

Digest AI · Research · published 2026-11-02T05:00:00Z

Canonical: https://digestai.news/story/gemma-3-12b-fails-to-detect-misinformation-when-poisoned-claims-domina

## Summary

Researchers found that the language model **Gemma-3-12b** struggles to identify misinformation when most of its retrieved documents support a false claim. The study, published on arXiv, tested 200 questions from the **SciFact** dataset, comparing scenarios where one document was poisoned against six healthy ones, versus six poisoned documents with one healthy one.

The model’s ability to detect misinformation dropped sharply: it correctly flagged only **25%** of poisoned claims in the minority condition (50/200) but just **10%** (20/200) when poisoned claims dominated. Of the 38 questions with differing outcomes, **34 shifted from detected to undetected**, while only four reversed. The researchers call this a ‘conformity bias’—the model conforms to the majority of retrieved evidence, even when it contradicts facts. The study highlights a critical flaw in **retrieval-augmented generation (RAG)**, where models may prioritize consensus over accuracy.

## Key points

- Gemma-3-12b’s misinformation detection drops from 25% to 10% when poisoned claims dominate retrieved evidence
- 34 of 38 questions shifted from detected to undetected in majority-poisoned scenarios, per the study’s analysis
- Researchers attribute the trend to a ‘conformity bias’ in RAG-based models, per arXiv paper

## Why it matters

This finding exposes a systemic risk in RAG systems, where models may ignore false claims if they appear credible due to sheer volume. It could force developers to rethink how retrieved evidence is weighted or validated before generation.

## Sources

1. [Language Models Show Conformity Bias in RAG: When the Majority Endorses a Poisoned Claim](https://arxiv.org/abs/2610.41012) (arXiv cs.CL, 2026-11-02, primary source)

## Cite

Digest AI, "Gemma-3-12b fails to detect misinformation when poisoned claims dominate retrieved evidence", 2 November 2026, https://digestai.news/story/gemma-3-12b-fails-to-detect-misinformation-when-poisoned-claims-domina

---

Written by Digest AI's editorial model from the linked sources; the sources are the record. Headlines, digests and key points are written by Digest AI and may be quoted with a link to the story page. Linked articles belong to their publishers. Terms: https://digestai.news/terms#reuse
JSON: https://digestai.news/story/gemma-3-12b-fails-to-detect-misinformation-when-poisoned-claims-domina.json