Knowledge conflicts for llms: A survey

representative citing papers

citing papers explorer

Showing 17 of 17 citing papers.

• LibEvoBench: Probing Temporal Knowledge Stratification in Code Generation Models cs.SE · 2026-06-24 · unverdicted · none · ref 50

  LibEvoBench benchmark shows LLMs are version-oblivious on evolving APIs, with documentation helping but version specification not.

• Predicting Future Behaviors in Reasoning Models Enables Better Steering cs.LG · 2026-06-09 · unverdicted · none · ref 24

  Probes predicting future behaviors from intermediate steps enable Future Probe Controlled Generation for steering large reasoning models with minimal quality degradation.

• TRACER: Token ReAssignment for Concept ERasure in Generative Recommendation cs.IR · 2026-06-05 · unverdicted · none · ref 52

  TRACER uses token reassignment for concept-related items plus a coherence regularizer to unlearn specific concepts in generative recommendation while preserving utility better than baselines.

• RAGognizer: Hallucination-Aware Fine-Tuning via Detection Head Integration cs.CL · 2026-04-17 · unverdicted · none · ref 4

  RAGognizer adds a detection head to LLMs for joint training on generation and token-level hallucination detection, yielding SOTA detection and fewer hallucinations in RAG while preserving output quality.

• OR-VSKC: Resolving Visual-Semantic Knowledge Conflicts in Operating Rooms with Synthetic Data-Guided Alignment cs.CV · 2025-06-25 · conditional · none · ref 29

  OR-VSKC provides 28,190 synthetic operating room images plus an expert subset to expose and reduce visual-semantic knowledge conflicts in multimodal models for surgical risk detection.

• SHIFT: Gate-Modulated Activation Steering for Knowledge Conflict Mitigation in Retrieval-Augmented Generation cs.CL · 2026-06-26 · unverdicted · none · ref 77

  SHIFT reformulates neuron editing as learnable gate modulation on under 0.01% parameters to let LLMs adaptively balance contextual and parametric knowledge during RAG generation.

• Trust, but Don't Verify: Epistemic Blind Spots in LLM Source Evaluation cs.LG · 2026-06-03 · unverdicted · none · ref 20

  LLMs identify fabricated statistics in isolation (rates 0.76-1.00) but ignore numeric validity during synthesis, relying on a methodology-register representation that transfers across domains.

• Persona Conditioning of Brand Recommendations in Retrieval-Augmented Commercial Chat: A Prominence-Stratified Cross-Provider Audit cs.AI · 2026-05-28 · unverdicted · none · ref 24

  Persona prefixes reduce brand recommendation Jaccard similarity by 0.12-0.20, with mid-market brands swapping up to 75% of recommendations while category leaders remain ~80% consistent across OpenAI and Anthropic models.

• Detecting Is Not Resolving: The Monitoring Control Gap in Retrieval Augmented LLMs cs.AI · 2026-05-26 · unverdicted · none · ref 28

  RAG models exhibit a monitoring-control gap: they acknowledge epistemic conflicts in accumulating documents yet fail to constrain unsafe recommendations, with single-turn tests overestimating safety.

• Three Regimes of Context-Parametric Conflict: A Predictive Framework and Empirical Validation cs.CL · 2026-05-12 · conditional · none · ref 21

  A three-regime framework resolves contradictions in LLM context vs. parametric knowledge conflicts by distinguishing single-source updating, competitive integration, and task-appropriate selection, with empirical confirmation of certainty gradients and task effects across five models.

• The Override Gap: A Magnitude Account of Knowledge Conflict Failure in Hypernetwork-Based Instant LLM Adaptation cs.LG · 2026-04-26 · conditional · none · ref 37 · 2 links

  Knowledge conflicts in hypernetwork LLM adaptation stem from constant adapter margins losing to frequency-dependent pretrained margins; selective layer boosting and conflict-aware triggering raise deep-conflict accuracy to 71-72.5% on Gemma-2B and Mistral-7B.

• ConflictRAG: Detecting and Resolving Knowledge Conflicts in Retrieval Augmented Generation cs.CL · 2026-05-17 · unverdicted · none · ref 4 · 2 links

  ConflictRAG introduces a conflict-aware RAG pipeline with two-stage detection (MLP + selective LLM), Entropy-TOPSIS credibility assessment, and a new CARS metric, reporting 88.7% F1 and 5.3-6.1% gains on benchmarks.

• MeMo: Memory as a Model cs.CL · 2026-05-14 · unverdicted · none · ref 4 · 2 links

  MeMo encodes new knowledge into a separate memory model that integrates with frozen LLMs, showing strong performance on QA benchmarks while avoiding catastrophic forgetting and working without access to model weights.

• Do LLMs have core beliefs? cs.LG · 2026-05-05 · unverdicted · none · ref 9

  LLMs generally fail to maintain stable worldviews under adversarial conversational pressure, indicating they lack core beliefs akin to those in human cognition.

• A Decomposition Perspective to Long-context Reasoning for LLMs cs.CL · 2026-04-09 · unverdicted · none · ref 20

  Decomposing long-context reasoning into atomic skills, synthesizing targeted pseudo-datasets, and applying RL improves LLM performance on long-context benchmarks by an average of 7.7%.

• Tug-of-War within A Decade: Conflict Resolution in Vulnerability Analysis via Teacher-Guided Retrieval-Augmented Generations cs.CL · 2026-03-25 · unverdicted · none · ref 31

  CRVA-TGRAG combines parent-document segmentation, ensemble retrieval, and teacher-guided fine-tuning to mitigate knowledge conflicts and improve accuracy in LLM-based CVE vulnerability analysis.

• Divergent Recommendations, Convergent Diagnoses: Cross-Provider Failure-Mode Convergence in AI Commercial Recommendation cs.CY · 2026-05-22 · unverdicted · none · ref 13

  Two major AI providers diverge in which brands they recommend but converge on classifying the failure reasons, especially for low-prominence brands.