Granuscore is a reference‑free metric that measures the granularity of text by exploiting the structure of a hierarchical embedding space. It successfully reproduces known hierarchical orderings on the Granola‑EQ dataset, distinguishes granularity across different discourse contexts, and explains sentence‑specificity variations beyond sentence length. The authors also apply Granuscore to four question‑answering benchmarks, revealing systematic differences in granularity among questions, gold answers, and model outputs, thereby offering a new lens for assessing QA dataset difficulty.
By Lukas Ellinger, Alexander Fichtl, Miriam Ansch\"utz, Georg Groh
The study examines how varying the context window size of GPT‑2 affects its alignment with human reading-time data. A U‑shaped relationship is found: very short windows (<20 tokens) capture local memory limits, while larger windows (500–1,000 tokens) provide the best overall fit. Counterfactual experiments that disrupt cross‑sentential coreference chains show that these long‑range coreference relations are crucial for the improved fit of larger context windows.
By Kohei Kajikawa, Lin Ai, Tatsuki Kuribayashi, Ethan Gotlieb Wilcox
arXiv:2607. 08284v1 Announce Type: new Abstract: Large language models (LLMs) have demonstrated rapidly improving long-context capabilities, prompting a wave of benchmarks designed to evaluate them.
By Siddhartha Jain, Ameya Velingker
The paper introduces LLM-Microscope, a toolkit for measuring how Large Language Models encode contextual information at the token level. It shows that seemingly minor tokens—such as determiners, stopwords, and punctuation—carry surprisingly high contextual weight, and removing them degrades performance on benchmarks like MMLU and BABILong-4k. The study also finds a strong link between contextualization and linearity, indicating that the transformation between layers can be approximated by a single linear mapping when tokens are well contextualized.
By Anton Razzhigaev, Matvey Mikhalchuk, Temurbek Rahmatullaev, Elizaveta Goncharova, Polina Druzhinina, Ivan Oseledets, Andrey Kuznetsov
The paper introduces the problem of cross‑lingual loopholes in large language model (LLM) unlearning, where forgetting a fact in one language can leave it accessible in others. It presents a new 174‑language benchmark, the Cross‑Lingual Unlearning Tensor, and proposes COVER, a method that selects a subset of source languages to maximize unlearning coverage under a language budget. Experiments show COVER reduces residual knowledge by 7.8–27.3% compared to uniform selection and works on both synthetic and real low‑resource news data.
By Tyler Skow, Shravan Chaudhari, Rama Chellappa, Abhay Yadav
Large language models (LLMs) have demonstrated rapidly improving long-context capabilities, prompting a wave of benchmarks designed to evaluate them. However, existing long-context evaluations - from Needle-in-a-Haystack (NIAH) tests to more recent multi-hop reasoning and summarization tasks - predominantly measure average-case performance, and many are either saturated or lack robustness.