Classical and neural NLP: translation, question answering, tokenization and the evaluation of language understanding.
arXiv:2608.21133v1 Announce Type: new Abstract: Medical image-text data can expose protected health information (PHI) through both visible image content as well as accompanying text, creating a barri...
The paper surveys 61 studies on mental‑health AI and identifies a misalignment in how trust is evaluated across disciplines. It proposes a three‑layer framework—human‑oriented, interaction‑oriented, and AI‑oriented trust—and maps stakeholder perspectives onto these layers. The authors argue that future research should focus on calibrating human trust to actual interaction and AI trustworthiness rather than merely maximizing perceived trust.
The paper presents an end‑to‑end framework for extracting and clustering trilingual Sri Lankan parliamentary debates in Sinhala, Tamil, and English. Using LLM‑based text extraction, multilingual embeddings, and density‑based clustering, the authors recover 30 macro‑topics with a cluster purity of 0.673. The temporal patterns of these topics align with major national events such as the 2019 Easter attacks and the 2022 economic crisis, demonstrating the method’s effectiveness where traditional LDA fails.
arXiv:2608.21019v1 Announce Type: cross Abstract: Quantization is widely used to deploy large language models, but its effect on uncertainty behavior, such as confidence, margins, and abstention, is...
The PSK submission to the WMT 2026 Multilingual Instruction Shared Task employs a 3.35B‑parameter Tiny Aya Global model enhanced with three QLoRA adapters, each dedicated to a specific task: multilingual summarization, passage‑based question answering, and filtered standalone question answering. The summarization adapter is trained on multilingual document‑summary pairs, including scientific papers with author‑written abstracts, and outperforms a multitask adapter trained solely on organizer data on a held‑out split. For open question answering, results vary with answer length and evaluation method, prompting the submission of three systems that share the same context and summarization adapters but differ in their open‑QA adapters.
SAGE (Self-Adaptive Generative Execution) introduces a unified framework for integrating AI functions into SQL by defining three typed primitives—AI_SCALAR, AI_AGG, and AI_JOIN—that correspond to the relational roles of transforming rows, aggregating groups, and joining row pairs. The framework standardizes a confidence-gated execution interface and tailors physical strategies to each primitive’s shape, with AI_JOIN employing predicate analysis and a recipe card to select optimal execution plans. Evaluations across scalar, aggregate, and join workloads demonstrate that SAGE consistently improves execution quality and efficiency, achieving the best overall SemBench performance and dramatically reducing model calls in factorable joins.
arXiv:2608.21252v1 Announce Type: cross Abstract: Question answering (QA) over long, connected documents remains challenging because relevant evidence may span multiple entities and their relationshi...
arXiv:2608.20925v1 Announce Type: cross Abstract: Reference-based metrics remain the standard choice in machine translation evaluation, partly because quality estimation methods often correlate less...
The paper introduces XKV, a latent protocol that enables efficient communication between heterogeneous language models by translating a sharer's key‑value cache into a receiver's context. XKV overcomes limitations of prior methods by jointly pooling both caches, reconciling differing layer depths, and allowing each receiver position to retrieve its own residual in native KV geometry. Across 45 dataset‑model pairings, XKV outperforms previous protocols and text communication while using fewer parameters and achieving faster translation times.
The paper introduces Llama-Mobile, a framework that quantizes vision‑language models for efficient mobile deployment. It uses a quantization pipeline that generates training data from the model itself, eliminating the need for the original training setup, and employs a novel 2.7‑bit‑per‑parameter format optimized for Arm CPUs. Applying this method, the authors compress the Llama 3.2 11B Vision Instruct model to 3.7 GB with 8‑bit activations while maintaining strong performance on visual question answering tasks.
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.
The paper introduces TSS (Triple-Stream Stress probe), a diagnostic framework that splits text into lexical, morpho-syntactic, and psycholinguistic style channels to analyze mental health NLP classifiers. Across four English datasets, TSS uncovers a lexical interference effect where adding lexical features harms performance on human-labeled data but not on auto-labeled data, and proposes the Degree of Divergence (DoD) statistic to audit label-source bias. The study demonstrates that style features largely remain effective even after masking content words, emphasizing that shortcut learning is label-source specific rather than clinically relevant.
arXiv:2608.20964v1 Announce Type: cross Abstract: In this research, we introduce SAraBERT, an enhanced version of AraBERT which proposes inter-sentence transformer layers for extractive summarization...
StateSight is a new benchmark designed to isolate and evaluate the ability of vision‑language models to reconstruct latent spatial structure from a single image. It consists of three procedurally generated task families—cube‑net opposite‑face reasoning, occluded cube‑tower counting, and 4‑neighbor connected‑component counting—each with 300 deterministic prompts and exact‑match scoring. The benchmark also includes a companion dataset, StateSight‑Steps, with 900 image‑text examples and 3,600 intermediate visual states to aid analysis of reconstruction errors.
SCOPE is a training‑free generative prompt‑compression framework that reduces LLM input length by chunking a prompt into semantically coherent segments, rewriting each chunk to be more concise, and then reconstructing a coherent prompt. Unlike token‑removal methods, SCOPE’s chunk‑level rewriting preserves critical information and text coherence, and includes optimization techniques for finer‑grained control of compression ratios. Extensive evaluations on question‑answering and summarization tasks show that SCOPE consistently outperforms selective compression baselines, especially at high compression ratios.
The paper introduces STAR-OPD, a structured aspect‑cascade‑aware on‑policy reward distillation method for aspect‑based sentiment analysis (ABSA) quadruple extraction. It addresses a specific failure mode where distilled models produce structurally invalid target‑aspect bindings, leading to hallucinated targets and corrupted downstream predictions. By training on student rollouts and applying set‑structured rewards that enforce binding consistency, target grounding, and fine‑grained aspect disambiguation, STAR‑OPD outperforms both off‑policy and generic on‑policy baselines on E‑ABSA20K and SemEval‑2014, reducing target hallucination and improving performance on structurally hard cases.
arXiv:2608.20886v1 Announce Type: cross Abstract: Real-world image search queries are multimodal and compositional: ``find this shirt in pink'' specifies an entity to retain, an attribute to modify,...
arXiv:2608. 20204v1 Announce Type: new Abstract: Legal work, with its heavy reliance on processing large amounts of text, is often considered one of the domains most exposed to the use of LLMs.
arXiv:2608. 18752v2 Announce Type: replace-cross Abstract: Statutory retrieval is necessary for citation-grounded legal question answering, but remains underexplored for Greek.
arXiv:2608. 11922v2 Announce Type: replace-cross Abstract: Predictive-distribution entropy is a strong answer-selection rule in retrieval-augmented generation (RAG) for question answering: across five QA benchmarks, selecting the answer a frozen respondent LLM produces with the lowest answer-token entropy lifts mean $F_1$ from 0.