Discourse-Role Labels as Presentation-Time Variables for Context Use in Language Models

• URL: http://arxiv.org/abs/2606.04109v2
• Date: Mon, 08 Jun 2026 04:40:12 GMT
• Title: Discourse-Role Labels as Presentation-Time Variables for Context Use in Language Models
• Abstract summary: We study the effect of Context-augmented language model systems on reader-model behavior.<n>We introduce a paired fixed-content probe over 500 MMLU-Pro items.<n>Each item receives the same misleading answer-bearing assertion under different discourse-role labels.<n>Across GPT-5.5, DeepSeek V4 Pro, Llama-3-8B-Instruct, and Qwen2.5-7B-Instruct, Misleading Adoption Rate shifts by 56-84 percentage points.
• Score: 3.376024929309146
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Context-augmented language model systems often wrap supplied content with labels such as Reference:, Evidence:, Instruction:, Note:, or Example:, but the effect of these labels on reader-model behavior remains underexplored. We introduce a paired fixed-content probe over 500 MMLU-Pro items: each item receives the same misleading answer-bearing assertion under different discourse-role labels, and adoption is measured by whether the model outputs the injected wrong option. Across GPT-5.5, DeepSeek V4 Pro, Llama-3-8B-Instruct, and Qwen2.5-7B-Instruct, Misleading Adoption Rate shifts by 56-84 percentage points. Binding or source-like labels such as Instruction: and Reference: produce high adoption, whereas Example: consistently suppresses it. Paired tests, bootstrap intervals, final-instruction ablations, and Qwen final-step log-probability probes support a label-conditioned candidate preference. Boundary probes show where the effect weakens or persists: arithmetic tasks reduce adoption, passage-shaped external context preserves smaller label gaps, short-answer evaluation rules out option-letter copying, and nested-label conflicts suggest that illustrative framing can delimit adoption scope. A 200-case single-author manual audit confirms that the short-answer contrasts are stable under conservative adjudication. The resulting claim is bounded but practical: context-utilization and reader-side RAG benchmarks should report and control wrapper labels, because presentation choices can change measured reliance on supplied context.

Related papers

• Explicit State Elicitation Is Not Enough: A Controlled Audit of Memory-Policy Classification [11.991995412984503]
  We use this setting to develop an empirical audit protocol for structured intermediate outputs.<n>A 480-example synthetic development set initially suggested large gains from a state-structured prompt bundle.<n>On this set, exposing the four state definitions improves accuracy, but an isolated explicit state-output field does not significantly improve policy accuracy.
  arXiv  Detail & Related papers  (2026-08-18T01:04:45Z)
• NL-PAC: Specification Ambiguity and Certified Minimax Risk Floors in LLM-Mediated Supervision [0.9475982252982437]
  We introduce Natural Language PAC, a framework that uses a fixed model's thresholded decoding law to define admissible labels and candidate targets.<n>In a frozen Qwen2.5--3B audit, one prespecified prompt yields a positive model-relative certificate, whereas a paraphrase and exact-rule controls yield zero.
  arXiv  Detail & Related papers  (2026-07-09T21:53:00Z)
• Quality and Agreement in Multilabel Emotion Annotation: A Case Study and Evaluation Framework [0.042970700836450486]
  We present a multilabel emotion annotation case study.<n>We examine how annotator behavior and aggregation choices affect both agreement estimates and downstream emotion classifiers.<n>Our results provide practical guidance for designing, aggregating, and evaluating multilabel emotion datasets.
  arXiv  Detail & Related papers  (2026-06-19T03:36:05Z)
• Multi-Label Test-Time Adaptation with Bayesian Conditional Priors [15.878958356965159]
  Conditional Priors Estimation (BCP) is a gradient-free test-time adaptation method that injects label dependency without tuning the backbone.<n>BCP operates without target annotations by estimating anchor-conditioned priors online from the unlabeled test stream.<n>BCP consistently outperforms strong TTA baselines across standard multi-label benchmarks and multiple CLIP backbones.
  arXiv  Detail & Related papers  (2026-06-11T05:29:00Z)
• Metric-Dependent Annotation Saturation for Learning from Label Distributions [0.0]
  We fine-tune NLI models on label distributions subsampled from ChaosNLI.<n>In our 3-class NLI setting, entropy correlation requires N 20-50 annotators to converge.<n>Soft labels carry item-specific signal that label smoothing cannot replicate.
  arXiv  Detail & Related papers  (2026-05-28T11:46:51Z)
• Human Label Variation as Stable Signal: Learning Annotator-Specific Explanation Behavior via Cross-Annotator Preference Optimization [51.93456979139756]
  Free-text explanations extend human label variation beyond label disagreement.<n>We study whether large language models can learn and reproduce such annotator-specific label-explanation behavior.
  arXiv  Detail & Related papers  (2026-05-27T17:55:00Z)
• RAPT: Retrieval-Augmented Post-hoc Thresholding for Multi-Label Classification [1.2599533416395765]
  We present RAPT, a deployment-oriented retrieval-augmented score thresholding wrapper.<n>RAPT is a model-agnostic wrapper for document representations for similarity search and per label confidence scores.<n>In industrial setting, RAPT achieved its best predictive performance with metric learners, reaching 0.87 Macro-F1.
  arXiv  Detail & Related papers  (2026-05-15T18:31:04Z)
• SpatialBench-UC: Uncertainty-Aware Evaluation of Spatial Prompt Following in Text-to-Image Generation [0.0]
  SpatialBench-UC is a small, reproducible benchmark for pairwise spatial relations.<n>We release a benchmark package, versioned prompts, pinned configs, per-sample checker outputs, and report tables.<n>We evaluate three baselines, Stable Diffusion 1.5, SD 1.5 BoxDiff, and SD 1.4 GLIGEN.
  arXiv  Detail & Related papers  (2026-01-19T23:37:10Z)
• Humans Hallucinate Too: Language Models Identify and Correct Subjective Annotation Errors With Label-in-a-Haystack Prompts [41.162545164426085]
  We explore label verification in contexts using Large Language Models (LLMs)<n>We propose the Label-in-a-Haystack Rectification (LiaHR) framework for subjective label correction.<n>This approach can be integrated into annotation pipelines to enhance signal-to-noise ratios.
  arXiv  Detail & Related papers  (2025-05-22T18:55:22Z)
• Mixed Blessing: Class-Wise Embedding guided Instance-Dependent Partial Label Learning [53.64180787439527]
  In partial label learning (PLL), every sample is associated with a candidate label set comprising the ground-truth label and several noisy labels.<n>For the first time, we create class-wise embeddings for each sample, which allow us to explore the relationship of instance-dependent noisy labels.<n>To reduce the high label ambiguity, we introduce the concept of class prototypes containing global feature information.
  arXiv  Detail & Related papers  (2024-12-06T13:25:39Z)
• Dist-PU: Positive-Unlabeled Learning from a Label Distribution Perspective [89.5370481649529]
  We propose a label distribution perspective for PU learning in this paper. Motivated by this, we propose to pursue the label distribution consistency between predicted and ground-truth label distributions. Experiments on three benchmark datasets validate the effectiveness of the proposed method.
  arXiv  Detail & Related papers  (2022-12-06T07:38:29Z)
• Category-Adaptive Label Discovery and Noise Rejection for Multi-label Image Recognition with Partial Positive Labels [78.88007892742438]
  Training multi-label models with partial positive labels (MLR-PPL) attracts increasing attention. Previous works regard unknown labels as negative and adopt traditional MLR algorithms. We propose to explore semantic correlation among different images to facilitate the MLR-PPL task.
  arXiv  Detail & Related papers  (2022-11-15T02:11:20Z)
• Semi-supervised Object Detection via Virtual Category Learning [68.26956850996976]
  This paper proposes to use confusing samples proactively without label correction. Specifically, a virtual category (VC) is assigned to each confusing sample. It is attributed to specifying the embedding distance between the training sample and the virtual category.
  arXiv  Detail & Related papers  (2022-07-07T16:59:53Z)
• Few-shot Slot Tagging with Collapsed Dependency Transfer and Label-enhanced Task-adaptive Projection Network [61.94394163309688]
  We propose a Label-enhanced Task-Adaptive Projection Network (L-TapNet) based on the state-of-the-art few-shot classification model -- TapNet. Experimental results show that our model significantly outperforms the strongest few-shot learning baseline by 14.64 F1 scores in the one-shot setting.
  arXiv  Detail & Related papers  (2020-06-10T07:50:44Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.