SinkTrack: Attention Sink based Context Anchoring for Large Language Models

• URL: http://arxiv.org/abs/2604.10027v1
• Date: Sat, 11 Apr 2026 04:49:11 GMT
• Title: SinkTrack: Attention Sink based Context Anchoring for Large Language Models
• Abstract summary: Large language models (LLMs) suffer from hallucination and context forgetting.<n>SinkTrack treats BOS> as an information anchor and injects key contextual features into its representation.<n>SinkTrack is training-free, plug-and-play, and introduces negligible inference overhead.
• Score: 57.87121346197518
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Large language models (LLMs) suffer from hallucination and context forgetting. Prior studies suggest that attention drift is a primary cause of these problems, where LLMs' focus shifts towards newly generated tokens and away from the initial input context. To counteract this, we make use of a related, intrinsic characteristic of LLMs: attention sink -- the tendency to consistently allocate high attention to the very first token (i.e., <BOS>) of a sequence. Concretely, we propose an advanced context anchoring method, SinkTrack, which treats <BOS> as an information anchor and injects key contextual features (such as those derived from the input image or instruction) into its representation. As such, LLM remains anchored to the initial input context throughout the entire generation process. SinkTrack is training-free, plug-and-play, and introduces negligible inference overhead. Experiments demonstrate that SinkTrack mitigates hallucination and context forgetting across both textual (e.g., +21.6% on SQuAD2.0 with Llama3.1-8B-Instruct) and multi-modal (e.g., +22.8% on M3CoT with Qwen2.5-VL-7B-Instruct) tasks. Its consistent gains across different architectures and scales underscore the robustness and generalizability. We also analyze its underlying working mechanism from the perspective of information delivery. Our source code is available at https://github.com/67L1/SinkTrack.

Related papers

• Stem: Rethinking Causal Information Flow in Sparse Attention [28.46760179998361]
  We propose Stem, a novel, plug-and-play sparsity module aligned with information flow.<n>First, Stem employs the Token Position-Decay strategy, applying position-dependent top-k within each layer to retain initial tokens.<n>Second, to preserve information-rich tokens, Stem utilizes the Output-Aware Metric.
  arXiv  Detail & Related papers  (2026-03-06T13:33:29Z)
• Step-Level Sparse Autoencoder for Reasoning Process Interpretation [48.99201531966593]
  Large Language Models (LLMs) have achieved strong complex reasoning capabilities through Chain-of-Thought (CoT) reasoning.<n>We propose step-level sparse autoencoder (SSAE), which serves as an analytical tool to disentangle different aspects of LLMs' reasoning steps into sparse features.<n> Experiments on multiple base models and reasoning tasks show the effectiveness of the extracted features.
  arXiv  Detail & Related papers  (2026-03-03T14:25:02Z)
• VCoT-Grasp: Grasp Foundation Models with Visual Chain-of-Thought Reasoning for Language-driven Grasp Generation [67.98487725287835]
  VCoT-Grasp is an end-to-end grasp foundation model that incorporates visual chain-of-thought reasoning to enhance visual understanding for grasp generation.<n>For training, we refine and introduce a large-scale dataset, VCoT-GraspSet, comprising 167K synthetic images with over 1.36M grasps.<n>Our method significantly improves grasp success rates and generalizes effectively to unseen objects, backgrounds, and distractors.
  arXiv  Detail & Related papers  (2025-10-07T11:50:26Z)
• CTR-Sink: Attention Sink for Language Models in Click-Through Rate Prediction [42.92011330807996]
  $textitCTR-Sink$ is a novel framework introducing behavior-level attention sinks tailored for recommendation scenarios.<n>Inspired by attention sink theory, it constructs attention focus sinks and dynamically regulates attention aggregation via external information.
  arXiv  Detail & Related papers  (2025-08-05T17:30:34Z)
• Enhancing Input-Label Mapping in In-Context Learning with Contrastive Decoding [71.01099784480597]
  Large language models (LLMs) excel at a range of tasks through in-context learning (ICL)<n>We introduce In-Context Contrastive Decoding (ICCD), a novel method that emphasizes input-label mapping.
  arXiv  Detail & Related papers  (2025-02-19T14:04:46Z)
• PAINT: Paying Attention to INformed Tokens to Mitigate Hallucination in Large Vision-Language Model [0.0]
  hallucinations often arise from the progressive weakening of attention weight to visual tokens.<n>textbfPAINT (textbfPaying textbfAttention to textbfINformed textbfTokens) is a plug-and-play framework that intervenes in the self-attention mechanism of the Large Vision Language Models.
  arXiv  Detail & Related papers  (2025-01-21T15:22:31Z)
• Looking Beyond Text: Reducing Language bias in Large Vision-Language Models via Multimodal Dual-Attention and Soft-Image Guidance [67.26434607115392]
  Large vision-language models (LVLMs) have achieved impressive results in various vision-language tasks. LVLMs suffer from hallucinations caused by language bias, leading to diminished focus on images and ineffective visual comprehension. We propose LACING to address the language bias of LVLMs with muLtimodal duAl-attention meChanIsm (MDA) aNd soft-image Guidance (IFG)
  arXiv  Detail & Related papers  (2024-11-21T16:33:30Z)
• When Attention Sink Emerges in Language Models: An Empirical View [39.36282162213973]
  Language Models (LMs) assign significant attention to the first token, even if it is not semantically important.<n>This phenomenon has been widely adopted in applications such as streaming/long context generation, KV cache optimization, inference acceleration, model quantization, and others.<n>We first demonstrate that attention sinks exist universally in LMs with various inputs, even in small models.
  arXiv  Detail & Related papers  (2024-10-14T17:50:28Z)
• SLAck: Semantic, Location, and Appearance Aware Open-Vocabulary Tracking [89.43370214059955]
  Open-vocabulary Multiple Object Tracking (MOT) aims to generalize trackers to novel categories not in the training set. We present a unified framework that jointly considers semantics, location, and appearance priors in the early steps of association. Our method eliminates complex post-processings for fusing different cues and boosts the association performance significantly for large-scale open-vocabulary tracking.
  arXiv  Detail & Related papers  (2024-09-17T14:36:58Z)
• StoryGPT-V: Large Language Models as Consistent Story Visualizers [33.68157535461168]
  generative models have demonstrated impressive capabilities in generating realistic and visually pleasing images grounded on textual prompts.<n>Yet, the emerging Large Language Model (LLM) showcases robust reasoning abilities to navigate through ambiguous references.<n>We introduce emphStoryGPT-V, which leverages the merits of the latent diffusion (LDM) and LLM to produce images with consistent and high-quality characters.
  arXiv  Detail & Related papers  (2023-12-04T18:14:29Z)

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.