# Multi-Channel Mitigation of Source-Trust Shortcuts in Fact-Checking RL Agents

arXiv: https://arxiv.org/abs/2609.36611 · HTML: https://mlsanity.com/paper/2609.36611/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 29 of its 34 references that are in the map.
- First posted on arXiv 29 Sep 2026.

## Placed in

- Retrieval-augmented generation

## Builds on (most cited first, 29 of 29)

- [DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models](https://mlsanity.com/paper/2402.03300/) (2024)
- [Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks](https://mlsanity.com/paper/2005.11401/) (2020)
- [On Calibration of Modern Neural Networks](https://mlsanity.com/paper/1706.04599/) (2017)
- [Shortcut Learning in Deep Neural Networks](https://mlsanity.com/paper/2004.07780/) (2020)
- [Self-RAG: Learning to Retrieve, Generate, and Critique through Self-Reflection](https://mlsanity.com/paper/2310.11511/) (2023)
- [Search-R1: Training LLMs to Reason and Leverage Search Engines with Reinforcement Learning](https://mlsanity.com/paper/2503.09516/) (2025)
- [Not what you've signed up for: Compromising Real-World LLM-Integrated Applications with Indirect Prompt Injection](https://mlsanity.com/paper/2302.12173/) (2023)
- [FEVER: a large-scale dataset for Fact Extraction and VERification](https://mlsanity.com/paper/1803.05355/) (2018)
- [Beyond Accuracy: Behavioral Testing of NLP models with CheckList](https://mlsanity.com/paper/2005.04118/) (2020)
- [Right for the Wrong Reasons: Diagnosing Syntactic Heuristics in Natural Language Inference](https://mlsanity.com/paper/1902.01007/) (2019)
- [Deep Reinforcement Learning at the Edge of the Statistical Precipice](https://mlsanity.com/paper/2108.13264/) (2021)
- [Deep Reinforcement Learning that Matters](https://mlsanity.com/paper/1709.06560/) (2017)
- [Annotation Artifacts in Natural Language Inference Data](https://mlsanity.com/paper/1803.02324/) (2018)
- [Fact or Fiction: Verifying Scientific Claims](https://mlsanity.com/paper/2004.14974/) (2020)
- [Learning the Difference that Makes a Difference with Counterfactually-Augmented Data](https://mlsanity.com/paper/1909.12434/) (2019)
- [A Survey on Automated Fact-Checking](https://mlsanity.com/paper/2108.11896/) (2021)
- ["Liar, Liar Pants on Fire": A New Benchmark Dataset for Fake News Detection](https://mlsanity.com/paper/1705.00648/) (2017)
- [SelectiveNet: A Deep Neural Network with an Integrated Reject Option](https://mlsanity.com/paper/1901.09192/) (2019)
- [Adaptive Chameleon or Stubborn Sloth: Revealing the Behavior of Large Language Models in Knowledge Conflicts](https://mlsanity.com/paper/2305.13300/) (2023)
- [Knowledge Conflicts for LLMs: A Survey](https://mlsanity.com/paper/2403.08319/) (2024)
- [HoVer: A Dataset for Many-Hop Fact Extraction And Claim Verification](https://mlsanity.com/paper/2011.03088/) (2020)
- [Accounting for Variance in Machine Learning Benchmarks](https://mlsanity.com/paper/2103.03098/) (2021)
- [FEVEROUS: Fact Extraction and VERification Over Unstructured and Structured information](https://mlsanity.com/paper/2106.05707/) (2021)
- [Counterfactual Fairness in Text Classification through Robustness](https://mlsanity.com/paper/1809.10610/) (2018)
- [AVeriTeC: A Dataset for Real-world Claim Verification with Evidence from the Web](https://mlsanity.com/paper/2305.13117/) (2023)
- [CrAM: Credibility-Aware Attention Modification in LLMs for Combating Misinformation in RAG](https://mlsanity.com/paper/2406.11497/) (2024)
- [Not All Contexts Are Equal: Teaching LLMs Credibility-aware Generation](https://mlsanity.com/paper/2404.06809/) (2024)
- [Resolving Conflicting Evidence in Automated Fact-Checking: A Study on Retrieval-Augmented LLMs](https://mlsanity.com/paper/2505.17762/) (2025)
- [Veri-R1: Toward Precise and Faithful Claim Verification via Online Reinforcement Learning](https://mlsanity.com/paper/2510.01932/) (2025)

## Abstract (arXiv)

Retrieval-augmented fact-checkers often receive a reliability label, such as HIGH or LOW trust, for each evidence source. These labels should adjust the model's confidence and its decision to search for more evidence, while the verdict should follow the evidence content. We introduce TrustSwap, a counterfactual test that swaps, lowers, or removes source labels while keeping every evidence text fixed, and measures its three output channels (the verdict, the confidence, and the search decision) separately. Across untrained and RL-trained models at two scales, three datasets, and two prompts, confidence and search respond to the labels as intended in 49 of 50 comparisons, yet a label change alone alters 4-23% of confident verdicts for Qwen3 models and up to 50% for an existing RL-trained fact-checker. Standard GRPO fine-tuning amplifies this shortcut at 8B in all six settings. To reduce it, we propose trust-swap augmentation (TSA), which trains GRPO on each claim with both its original and its label-swapped evidence under the same gold verdict. At 4B, TSA lowers the verdict flip rate by 7-35% (relative) in four of six settings, keeps accuracy and the intended confidence and search responses, outperforms reward-based alternatives in the main setting, and carries over to an unseen label-removal perturbation. An added consistency reward helps on the trained-on swap but not on unseen perturbations. At 8B, TSA's effect is not detectable, which makes scale the main open question.