mlsanityData to 7 Oct 2026

The First Drop of Ink: Nonlinear Impact of Distracting Information in Long-Context Reasoning

arXiv 2605.10828 ↗First posted 11 May 2026cs.AIcited by 0

On the map: LLM reasoning → Reasoning benchmarks and evaluation

also in 2 more
Abstract (arXiv)

As large language models are increasingly deployed in retrieval-augmented generation and agentic systems that accumulate extensive context, understanding how distracting information affects long-context performance becomes critical. Prior work shows that semantically relevant yet misleading documents degrade performance, but the quantitative relationship between the proportion of distractors and performance remains unstudied. In this work, we systematically vary the hard-distractor proportion in fixed-length contexts, revealing a striking nonlinear pattern: as the proportion of hard distractors increases, performance drops sharply within the first small fraction, while the remainder of the range yields only marginal additional decline. We term this ''The First Drop of Ink'' effect, analogous to how a single drop of ink contaminates water. Our theoretical and empirical analyses grounded in attention mechanics show that hard distractors capture disproportionate attention even at small proportions, with diminishing marginal impact as their proportion grows. Controlled experiments further show that filtering gains mainly come from context-length reduction rather than distractor removal; substantial recovery requires reducing the hard-distractor proportion to near zero, highlighting the importance of upstream retrieval precision.

Key facts. No citing papers in the map yet (data through 7 Oct 2026). Placed in Reasoning benchmarks and evaluation, one of 17 directions in LLM reasoning. Builds on 9 of its 51 references that are in the map. First posted on arXiv 11 May 2026.

Builds on

9 of 51 references.

Closest ancestors · 1

Same direction, newest first

  1. 2026Lost in the Noise: How Reasoning Models Fail with Contextual Distractors12 Jan

Other references · 8

  1. 2024Gemini 1.5: Unlocking multimodal understanding across millions of tokens of contextcited by 3,175
  2. 2023AutoGen: Enabling Next-Gen LLM Applications via Multi-Agent Conversationcited by 1,906
  3. 2024∞Bench: Extending Long Context Evaluation Beyond 100K Tokenscited by 360
  4. 2025Agentic Retrieval-Augmented Generation: A Survey on Agentic RAGcited by 242
  5. 2025From Web Search towards Agentic Deep Research: Incentivizing Search with Reasoning Agentscited by 49
  6. 2025Hidden in the Haystack: Smaller Needles are More Difficult for LLMs to Findcited by 7
  7. 2025A Controllable Examination for Long-Context Language Modelscited by 7
  8. 2025Learning to Focus: Focal Attention for Selective and Scalable Transformerscited by 2

Built on by

Not cited yet.

Similar, not cited

Close in wording, no citation either way

  1. 2025Understanding the Effects of Distractors on Reasoning Vision-Language Models
  2. 2026Distractor-Aware Truncation: Disentangling Context-Length Effects from Signal Loss in Long-Context LLM Benchmarks
  3. 2026The Cost of Context: Mitigating Textual Bias in Multimodal Retrieval-Augmented Generation
  4. 2025Longer Context, Deeper Thinking: Uncovering the Role of Long-Context Ability in Reasoning
  5. 2026Lexical Perturbations Disrupt LLM Reasoning: An Empirical Study of Attention Diversion
  6. 2026Beyond Fine-Tuning: In-Context Learning and Chain-of-Thought for Reasoned Distractor Generation
  7. 2024Retrieval Head Mechanistically Explains Long-Context Factuality
  8. 2025When Can Large Reasoning Models Save Thinking? Mechanistic Analysis of Behavioral Divergence in Reasoning