mlsanityData to 7 Oct 2026

Reinforcement Learning for Large Language Model Selective Evidence Adoption from Contaminated Retrieval Results

arXiv 2607.20090 ↗First posted 22 Jul 2026cs.CL cs.AIcited by 2

On the map: Retrieval-augmented generation → Agentic RAG, agents and memory

Abstract (arXiv)

Retrieval-augmented large language models frequently face contexts that interleave useful evidence with misleading statements or instruction-like content. Blanket refusal discards valid evidence, whereas uncritical adoption yields incorrect or unsafe answers. The ability to selectively adopt relevant information while rejecting deceptive or harmful content is therefore critical for reliable deployment in real-world retrieval settings. We introduce SelectBench, a controlled benchmark and training set for selective evidence adoption, and post-train Qwen3.5-4B directly with DAPO using either deterministic rule rewards or a frozen semantic judge. On the corrected 325-example SelectBench-v2 test set, strict success rises from 22.46% for the original checkpoint to 25.54% with DAPO-Rule and 26.46% with DAPO-DeepSeek. Both trained policies reduce forbidden-content adoption and produce shorter, more focused responses, yet prompt-injection following does not improve. The paired gains are modest and fail to survive Holm correction, suggesting that stronger reward shaping or additional training iterations may be needed for more robust gains. DAPO-DeepSeek exhibits no material degradation on MMLU or clean HotpotQA, indicating that the post-training procedure preserves general capabilities. These results demonstrate a directional improvement in selective evidence use, while identifying injection resistance and statistical robustness as important remaining challenges for future work.

Key facts. Cited by 2 papers in the map (data through 7 Oct 2026). Placed in Agentic RAG, agents and memory, one of 15 directions in Retrieval-augmented generation. Builds on 23 of its 26 references that are in the map. First posted on arXiv 22 Jul 2026.

Builds on

23 of 26 references.

References · 23

  1. 2017Proximal Policy Optimization Algorithmscited by 14,441
  2. 2024DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Modelscited by 8,851
  3. 2023Direct Preference Optimization: Your Language Model is Secretly a Reward Modelcited by 7,916
  4. 2020Measuring Massive Multitask Language Understandingcited by 7,768
  5. 2018HotpotQA: A Dataset for Diverse, Explainable Multi-hop Question Answeringcited by 3,686
  6. 2025DAPO: An Open-Source LLM Reinforcement Learning System at Scalecited by 2,678
  7. 2020Constructing A Multi-hop QA Dataset for Comprehensive Evaluation of Reasoning Stepscited by 1,259
  8. 2021MuSiQue: Multihop Questions via Single-hop Question Compositioncited by 893
  9. 2026DeepSeek-V4: Towards Highly Efficient Million-Token Context Intelligencecited by 877
  10. 2025Group Sequence Policy Optimizationcited by 788
  11. 2025ReSearch: Learning to Reason with Search for LLMs via Reinforcement Learningcited by 152
  12. 2025FalseReject: A Resource for Improving Contextual Safety and Mitigating Over-Refusals in LLMs via Structured Reasoningcited by 53
11 more
  1. 2026Harness-Bench: Measuring Harness Effects across Models in Realistic Agent Workflowscited by 44
  2. 2026Are Tools All We Need? Unveiling the Tool-Use Tax in LLM Agentscited by 11
  3. 2025Unified attacks to large language model watermarks: spoofing and scrubbing in unauthorized knowledge distillationcited by 9
  4. 2025SEM: Reinforcement Learning for Search-Efficient Large Language Modelscited by 9
  5. 2025Latent-space adversarial training with post-aware calibration for defending large language models against jailbreak attackscited by 7
  6. 2025Unified Defense for Large Language Models against Jailbreak and Fine-Tuning Attacks in Educationcited by 5
  7. 2026A Survey on Evaluating Quality and Trustworthiness in LLM-Generated Datacited by 5
  8. 2025Hierarchical Safety Realignment: Lightweight Restoration of Safety in Pruned Large Vision-Language Modelscited by 4
  9. 2026SURGENT: A Surgical Multi-Agent Assistance System Across the Perioperative Workflowcited by 3
  10. 2025From Construction to Injection: Edit-Based Fingerprints for Large Language Modelscited by 2
  11. 2026AGMark: Attention-Guided Dynamic Watermarking for Large Vision-Language Modelscited by 1

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2026: 2.

2 papers.

  1. 2026EMR: Self-Evolving Medical Multi-Agent System via Experience Mining and Reusecited by 1
  2. 2026Evaluating LLMs in Database Scenarios: A Lifecycle Benchmark for Assessing Their Potential in Core Database Taskscited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2026OThink-SRR1: Search, Refine and Reasoning with Reinforced Learning for Large Language Models
  2. 2025Learning to Extract Rational Evidence via Reinforcement Learning for Retrieval-Augmented Generation
  3. 2026GRACE: Reinforcement Learning for Grounded Response and Abstention under Contextual Evidence
  4. 2026Reinforcement Inference: Leveraging Uncertainty for Self-Correcting Language Model Reasoning
  5. 2026Learning Evidence Highlighting for Frozen LLMs
  6. 2025Towards Transparent RAG: Fostering Evidence Traceability in LLM Generation via Reinforcement Learning
  7. 2025Rethinking On-policy Optimization for Query Augmentation
  8. 2026Evidence-State Rewards for Long-Context Reasoning