MUDDLE is a benchmark designed to disentangle the effects of document length and topical distractors on document question‑answering systems. It contains 270 human‑annotated questions, each tested in five conditions: the source alone, the source with two or four hard negatives (topically similar), and the source with two or four random distractors matched in length and provenance. Experiments with GPT‑5‑mini show that hard negatives reduce accuracy more than length‑matched random distractors, indicating that topical similarity is a more significant source of error than length alone.
arXiv:2605. 10828v2 Announce Type: replace Abstract: 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.
arXiv:2608. 03297v1 Announce Type: new Abstract: A standard claim in the literature on retrieval-augmented and memory-augmented language models is that shorter context is better when the relevant information is preserved.
arXiv:2609.22101v1 Announce Type: cross Abstract: Large language models can process increasingly long prompts, yet their ability to locate and use decisive evidence may degrade as irrelevant or confu...
The paper introduces the Controlled Attribution Benchmark (CATT), a synthetic evaluation framework that fine‑tunes language models on entirely fictitious knowledge to provide clear ground truth for data attribution. By systematically varying single dataset properties—such as document redundancy and lexical distractors—CATT reveals that these factors consistently degrade attribution performance. The study also explores two underexamined design choices: using negative log‑odds gradients instead of cross‑entropy gradients, and aggregating only the top token attribution scores, which together improve ground‑truth recall by 35–72 percentage points and enhance counterfactual retraining performance on FTRACE‑TREX by up to 4.9×.
arXiv:2609.35794v1 Announce Type: cross Abstract: Just as Socrates recognized the limits of his own knowledge, Retrieval-Augmented Language Models (RALMs) should learn to abstain when the retrieved e...
arXiv:2610.26951v1 Announce Type: new Abstract: Retrieval-augmented generation (RAG) is the dominant paradigm for grounding large language models in external corpora, but its similarity-driven, chunk...
arXiv:2602. 00344v2 Announce Type: replace-cross Abstract: While Retrieval-Augmented Generation (RAG) is one of the dominant paradigms for enhancing Large Vision-Language Models (LVLMs) on knowledge-based VQA tasks, recent work attributes RAG failures to insufficient attention towards the retrieved context, proposing to reduce the attention allocated to image tokens.
The paper introduces Highlight-Then-Summarize (H2S), a two-step approach that first highlights question-relevant evidence in long documents and then condenses it into a compact, question-conditioned summary before generating an answer. The authors built the H2S-Dataset with 6,647 examples spanning 11 benchmark families, and developed H2S-RL to reward evidence selection and summary construction. Evaluated on the H2S-Bench suite, the H2S-14B model outperforms larger open-source models, achieving the highest overall score and maintaining strong performance even with a reduced output budget.
arXiv:2607. 10400v1 Announce Type: cross Abstract: Vision language models (VLMs) have achieved strong performance on visual document understanding benchmarks such as DocVQA, ChartQA, and MMLongBench-Doc.
arXiv:2608. 04569v1 Announce Type: cross Abstract: Hard prompt compression reduces long-context inference cost by independently scoring tokens, sentences, or chunks and retaining the highest-scoring units under a budget.
arXiv:2608. 16185v1 Announce Type: cross Abstract: LLM agents increasingly answer questions over dynamic raw-document collections, where files may change before preprocessing, and relevant evidence (spans, sections, pages, or tables) is query-dependent.
The paper identifies that language models lose accuracy when prompts include non‑evidential context—irrelevant statements or random tokens that do not overlap semantically with the query. This degradation is traced to an "Intrusion Pathway" involving specific attention heads that amplify distractor content and weaken correct answers. By targeting and masking a small set of these intrusion heads, the study recovers most of the lost performance with minimal impact on useful context.