mlsanityData to 14 Oct 2026

LODESTAR: Robust Entropy-Based Answer Selection in Retrieval-Augmented Generation for Question Answering -- Directing Frozen-LLM Entropy with a Reinforcement-Learned Prompt Polarizer under Misleading Passages

arXiv 2608.11922 ↗First posted 12 Aug 2026cs.CL cs.IR cs.LGcited by 0

On the map: Retrieval-augmented generation → RAG robustness and adaptive retrieval

Abstract (arXiv)

Predictive-distribution entropy is a strong answer-selection rule in retrieval-augmented generation (RAG) for question answering: across five QA benchmarks, selecting the answer a frozen respondent LLM produces with the lowest answer-token entropy lifts mean F1 from 0.4769 to 0.5148 over the retriever's top-ranked passage, without gold answers. Yet this rule, which prior entropy-based selectors adopt, fails: a misleading passage makes the respondent confidently wrong, driving entropy down where the uncertainty signal looks most trustworthy. The failure comes from the passage the respondent reads, and the context it is read in is an input we can intervene on. We introduce LODESTAR: to our knowledge the first method to score a text intervention by the uncertainty it induces in a third-party frozen respondent, compared within one question. LODESTAR uses reinforcement learning (GRPO) to train, once and offline, a polarizer -- a short fixed natural-language string inserted into the respondent's prompt and never into its weights, directing entropy so that entropy-based answer selection stays robust to misleading passages; training labels are built from gold answers and two LLM judges, and inference reads neither. With every competing selector under the same frozen respondent and candidate pools on 5,008 questions, LODESTAR attains the highest mean F1 of any inference-ready selector (0.5339), the highest macro exact match (0.4136), and the highest GPT-4o judge score of the frozen-respondent configurations judged (0.6435); its three-seed mean wins all 70 F1 cells against fourteen published configurations and is paired-significant on F1 against every one. The gain holds in-domain on NQ-Open and out-of-domain over SQuAD, TriviaQA, EntityQuestions and WebQuestions. Ablating the polarizer shows it is what makes the respondent read a misleading passage less often (26.0% vs 30.3%).

Key facts. No citing papers in the map yet (data through 14 Oct 2026). Placed in RAG robustness and adaptive retrieval, one of 15 directions in Retrieval-augmented generation. Builds on 38 of its 49 references that are in the map. First posted on arXiv 12 Aug 2026.

Builds on

38 of 49 references.

Closest ancestors · 5

Same direction, newest first

  1. 2026CRITIC-R1: Learning Structured Critics for Retrieval-Augmented Generation28 May
  2. 2025GainRAG: Preference Alignment in Retrieval-Augmented Generation through Gain Signal Synthesis24 May
  3. 2025Adaptive Retrieval Without Self-Knowledge? Bringing Uncertainty Back Home22 Jan
  4. 2024SeaKR: Self-aware Knowledge Retrieval for Adaptive Retrieval Augmented Generation27 Jun
  5. 2023Self-RAG: Learning to Retrieve, Generate, and Critique through Self-Reflection17 Oct

Other references · 33

  1. 2024The Llama 3 Herd of Modelscited by 14,715
  2. 2024DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Modelscited by 9,379
  3. 2025Qwen3 Technical Reportcited by 8,820
  4. 2023Judging LLM-as-a-Judge with MT-Bench and Chatbot Arenacited by 8,626
  5. 2024Qwen2.5 Technical Reportcited by 4,942
  6. 2024GPT-4o System Cardcited by 4,588
  7. 2016SQuAD: 100,000+ Questions for Machine Comprehension of Textcited by 4,148
  8. 2017TriviaQA: A Large Scale Distantly Supervised Challenge Dataset for Reading Comprehensioncited by 2,955
  9. 2025DAPO: An Open-Source LLM Reinforcement Learning System at Scalecited by 2,810
  10. 2025gpt-oss-120b & gpt-oss-20b Model Cardcited by 1,386
  11. 2025Understanding R1-Zero-Like Training: A Critical Perspectivecited by 1,345
  12. 2025Qwen3 Embedding: Advancing Text Embedding and Reranking Through Foundation Modelscited by 1,261
21 more
  1. 2021Unsupervised Dense Information Retrieval with Contrastive Learningcited by 1,216
  2. 2024M3-Embedding: Multi-Linguality, Multi-Functionality, Multi-Granularity Text Embeddings Through Self-Knowledge Distillationcited by 1,073
  3. 2025GEPA: Reflective Prompt Evolution Can Outperform Reinforcement Learningcited by 542
  4. 2023Detecting Pretraining Data from Large Language Modelscited by 387
  5. 2024Replacing Judges with Juries: Evaluating LLM Generations with a Panel of Diverse Modelscited by 326
  6. 2024INSIDE: LLMs' Internal States Retain the Power of Hallucination Detectioncited by 323
  7. 2024Optimizing Instructions and Demonstrations for Multi-Stage Language Model Programscited by 291
  8. 2021Simple Entity-Centric Questions Challenge Dense Retrieverscited by 140
  9. 2023Large language models can accurately predict searcher preferencescited by 131
  10. 2024Min-K%++: Improved Baseline for Detecting Pre-Training Data from Large Language Modelscited by 103
  11. 2024FIRST: Faster Improved Listwise Reranking with Single Token Decodingcited by 54
  12. 2024MBA-RAG: a Bandit Approach for Adaptive Retrieval-Augmented Generation through Question Complexitycited by 29
  13. 2024Entropy-Based Decoding for Retrieval-Augmented Large Language Modelscited by 19
  14. 2025Semantic Energy: Detecting LLM Hallucination Beyond Entropycited by 18
  15. 2025Why Uncertainty Estimation Methods Fall Short in RAG: An Axiomatic Analysiscited by 14
  16. 2025Learned Hallucination Detection in Black-Box LLMs using Token-level Entropy Production Ratecited by 10
  17. 2025How Well Does First-Token Entropy Approximate Word Entropy as a Psycholinguistic Predictor?cited by 4
  18. 2026Less is More for RAG: Information Gain Pruning for Generator-Aligned Reranking and Evidence Selectioncited by 3
  19. 2026The Anatomy of Uncertainty in LLMscited by 2
  20. 2025Beyond Perplexity: Let the Reader Select Retrieval Summaries via Spectrum Projection Scorecited by 1
  21. 2026CTRL-RAG: Contrastive Likelihood Reward Based Reinforcement Learning for Context-Faithful RAG Modelscited by 1

Built on by

Not cited yet.

Similar, not cited

Close in wording, no citation either way

  1. 2026Rethinking the Reranker: Boundary-Aware Evidence Selection for Robust Retrieval-Augmented Generation
  2. 2025Retrieval as a Decision: Training-Free Adaptive Gating for Efficient RAG
  3. 2025SUGAR: Leveraging Contextual Confidence for Smarter Retrieval
  4. 2026Predicting the Benefit of Retrieval Augmentation in Open-Domain Question Answering
  5. 2026Beyond Self-Knowledge: Propagating Uncertainty Across Reasoning and Retrieval in LLMs
  6. 2023PRCA: Fitting Black-Box Large Language Models for Retrieval Question Answering via Pluggable Reward-Driven Contextual Adapter
  7. 2026Beyond the Context Window: An Adaptive Entropy-Based Routing Framework for Hybrid Retrieval and Long-Context Language Models
  8. 2026Interpretable Uncertainty for Adaptive Retrieval and Reasoning in Question Answering