How Context Attribution Handles What the Model Already Knows

• URL: http://arxiv.org/abs/2607.23804v1
• Date: Sun, 26 Jul 2026 19:06:29 GMT
• Title: How Context Attribution Handles What the Model Already Knows
• Abstract summary: Context attribution methods for large language models (LLMs) identify which input context contributes to the model response.<n>Recent works show the initial success in attributing the con-tributive score of the contexts.<n>We observe that when the context overlaps with the training data, these methods can- not disentangle in-context from in-weight contributions.
• Score: 12.541488378965932
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Context attribution methods for large language models (LLMs) identify which input context contributes to the model response. Recent works show the initial success in attributing the con- tributive score of the contexts. However, we observe that when the context overlaps with the training data, these methods can- not disentangle in-context from in-weight (IW) contributions, producing unreliable scores. Based on this observation, in this work, we introduce: 1) an evaluation protocol that relies on four new metrics (base-model context attribution score (BCS), cross-model context attribution consistency (CAC), attribution preservation score (APS), source separation pre- cision (SSP)) and 2) a benchmark dataset (WMDP-Cyber++) with ground-truth provenance labels to systematically assess attribution under IW overlap. In our experiments across four well-known context attribution methods, we demonstrate that they provide unfaithful attribution when the knowledge from the context also exists in the weights. Finally, we adapt these methods for source separation (IW vs. in-context learning (ICL)) and show that they cannot do the disentanglement based on the contributive score

Related papers

• Context is Enough: Empirical Validation of $\ extit{Sequentiality}$ on Essays [1.338174941551702]
  We show that the contextual version of sequentiality aligns more closely with human assessments of discourse-level traits.<n>Our findings support the use of context-based sequentiality as a validated, interpretable, and complementary feature for automated essay scoring and related NLP tasks.
  arXiv  Detail & Related papers  (2025-11-12T10:31:07Z)
• Evaluation Framework for Highlight Explanations of Context Utilisation in Language Models [36.64390220306208]
  Context utilisation is the ability of Language Models to incorporate relevant information from the provided context when generating responses.<n>We introduce the first gold standard HE evaluation framework for context attribution, using controlled test cases with known ground-truth context usage.<n>We find that MechLight performs best across all context scenarios.
  arXiv  Detail & Related papers  (2025-10-03T00:15:36Z)
• Label-Context-Dependent Internal Language Model Estimation for CTC [50.25063912757367]
  We propose novel context-dependent ILM estimation methods for connectionist temporal classification.<n> Experimental results show that context-dependent ILMs outperform the context-independent priors in cross-domain evaluation.<n>The proposed label-level KD with smoothing method surpasses other ILM estimation approaches.
  arXiv  Detail & Related papers  (2025-06-06T13:54:43Z)
• Document Attribution: Examining Citation Relationships using Large Language Models [62.46146670035751]
  We propose a zero-shot approach that frames attribution as a straightforward textual entailment task.<n>We also explore the role of the attention mechanism in enhancing the attribution process.
  arXiv  Detail & Related papers  (2025-05-09T04:40:11Z)
• Controlling Out-of-Domain Gaps in LLMs for Genre Classification and Generated Text Detection [0.20482269513546458]
  This study demonstrates that the modern generation of Large Language Models (LLMs) suffers from the same out-of-domain (OOD) performance gap observed in prior research on pre-trained Language Models (PLMs)<n>We introduce a method that controls which predictive indicators are used and which are excluded during classification.<n>This approach reduces the OOD gap by up to 20 percentage points in a few-shot setup.
  arXiv  Detail & Related papers  (2024-12-29T21:54:39Z)
• Context-DPO: Aligning Language Models for Context-Faithfulness [80.62221491884353]
  We propose the first alignment method specifically designed to enhance large language models' context-faithfulness.<n>By leveraging faithful and stubborn responses to questions with provided context from ConFiQA, our Context-DPO aligns LLMs through direct preference optimization.<n>Extensive experiments demonstrate that our Context-DPO significantly improves context-faithfulness, achieving 35% to 280% improvements on popular open-source models.
  arXiv  Detail & Related papers  (2024-12-18T04:08:18Z)
• On the Loss of Context-awareness in General Instruction Fine-tuning [101.03941308894191]
  We investigate the loss of context awareness after supervised fine-tuning.<n>We find that the performance decline is associated with a bias toward different roles learned during conversational instruction fine-tuning.<n>We propose a metric to identify context-dependent examples from general instruction fine-tuning datasets.
  arXiv  Detail & Related papers  (2024-11-05T00:16:01Z)
• On Context Utilization in Summarization with Large Language Models [83.84459732796302]
  Large language models (LLMs) excel in abstractive summarization tasks, delivering fluent and pertinent summaries. Recent advancements have extended their capabilities to handle long-input contexts, exceeding 100k tokens. We conduct the first comprehensive study on context utilization and position bias in summarization.
  arXiv  Detail & Related papers  (2023-10-16T16:45:12Z)
• Context vs Target Word: Quantifying Biases in Lexical Semantic Datasets [18.754562380068815]
  State-of-the-art contextualized models such as BERT use tasks such as WiC and WSD to evaluate their word-in-context representations. This study presents the first quantitative analysis (using probing baselines) on the context-word interaction being tested in major contextual lexical semantic tasks.
  arXiv  Detail & Related papers  (2021-12-13T15:37:05Z)
• An Explanation of In-context Learning as Implicit Bayesian Inference [117.19809377740188]
  We study the role of the pretraining distribution on the emergence of in-context learning. We prove that in-context learning occurs implicitly via Bayesian inference of the latent concept. We empirically find that scaling model size improves in-context accuracy even when the pretraining loss is the same.
  arXiv  Detail & Related papers  (2021-11-03T09:12:33Z)
• CIF-based Collaborative Decoding for End-to-end Contextual Speech Recognition [14.815422751109061]
  We propose a continuous integrate-and-fire (CIF) based model that supports contextual biasing in a more controllable fashion. An extra context processing network is introduced to extract contextual embeddings, integrate acoustically relevant context information and decode the contextual output distribution. Our method brings relative character error rate (CER) reduction of 8.83%/21.13% and relative named entity character error rate (NE-CER) reduction of 40.14%/51.50% when compared with a strong baseline.
  arXiv  Detail & Related papers  (2020-12-17T09:40:11Z)

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.