HFEPX Daily Archive: 2026-02-17

Research Utility Snapshot

Papers Published On This Date

• Updating Parametric Knowledge with Context Distillation Retains Post-Training Capabilities

  Shankar Padmanabhan, Mustafa Omer Gul, Tanya Goyal · Feb 17, 2026

  Post-training endows pretrained LLMs with a variety of desirable skills, including instruction-following, reasoning, and others.

• Why Any-Order Autoregressive Models Need Two-Stream Attention: A Structural-Semantic Tradeoff

  Patrick Pynadath, Ruqi Zhang · Feb 17, 2026

  Any-order autoregressive models (AO-ARMs) offer a promising path toward efficient masked diffusion by enabling native key-value caching, but competitive performance has so far required two-stream attention, typically motivated as a means of

• Language Statistics and False Belief Reasoning: Evidence from 41 Open-Weight LMs

  Sean Trott, Samuel Taylor, Cameron Jones, James A. Michaelov, Pamela D. Rivière · Feb 17, 2026

  Research on mental state reasoning in language models (LMs) has the potential to inform theories of human social cognition--such as the theory that mental state reasoning emerges in part from language exposure--and our understanding of LMs

• Surgical Activation Steering via Generative Causal Mediation

  Aruna Sankaranarayanan, Amir Zur, Atticus Geiger, Dylan Hadfield-Menell · Feb 17, 2026

  Where should we intervene in a language model (LM) to control behaviors that are diffused across many tokens of a long-form response?

• CLAA: Cross-Layer Attention Aggregation for Accelerating LLM Prefill

  Bradley McDanel, Steven Li, Harshit Khaitan · Feb 17, 2026

  This oracle reveals that existing heuristics exhibit high variance across layers: rankings can degrade sharply at specific layers, a failure mode invisible to end-to-end benchmarks.

• Multi-Objective Alignment of Language Models for Personalized Psychotherapy

  Mehrab Beikzadeh, Yasaman Asadollah Salmanpour, Ashima Suvarna, Sriram Sankararaman, Matteo Malgaroli · Feb 17, 2026

  Pairwise PreferenceExpert Verification

  While AI systems show therapeutic promise, current alignment approaches optimize objectives independently, failing to balance patient preferences with clinical safety.

• Evidence-Grounded Subspecialty Reasoning: Evaluating a Curated Clinical Intelligence Layer on the 2025 Endocrinology Board-Style Examination

  Amir Hosseinian, MohammadReza Zare Shahneh, Umer Mansoor, Gilbert Szeto, Kirill Karlin · Feb 17, 2026

  Results: Mirror achieved 87.5% accuracy (105/120; 95% CI: 80.4-92.3%), exceeding a human reference of 62.3% and frontier LLMs including GPT-5.2 (74.6%), GPT-5 (74.0%), and Gemini-3-Pro (69.8%).

• A Curious Class of Adpositional Multiword Expressions in Korean

  Junghyun Min, Na-Rae Han, Jena D. Hwang, Nathan Schneider · Feb 17, 2026

  Multiword expressions (MWEs) have been widely studied in cross-lingual annotation frameworks such as PARSEME.

• MAEB: Massive Audio Embedding Benchmark

  Adnan El Assadi, Isaac Chung, Chenghao Xiao, Roman Solomatin, Animesh Jha · Feb 17, 2026

  We introduce the Massive Audio Embedding Benchmark (MAEB), a large-scale benchmark covering 30 tasks across speech, music, environmental sounds, and cross-modal audio-text reasoning in 100+ languages.

• Anatomy of Capability Emergence: Scale-Invariant Representation Collapse and Top-Down Reorganization in Neural Networks

  Jayadev Billa · Feb 17, 2026

  Capability emergence during neural network training remains mechanistically opaque.

• DocSplit: A Comprehensive Benchmark Dataset and Evaluation Approach for Document Packet Recognition and Splitting

  Md Mofijul Islam, Md Sirajus Salekin, Nivedha Balakrishnan, Vincil C. Bishop, Niharika Jain · Feb 17, 2026

  We present the first comprehensive benchmark dataset, DocSplit, along with novel evaluation metrics for assessing the document packet splitting capabilities of large language models.

• Avey-B

  Devang Acharya, Mohammad Hammoud · Feb 17, 2026

  Results show that this reformulated architecture compares favorably to four widely used Transformer-based encoders, consistently outperforming them on standard token-classification and information-retrieval benchmarks while scaling more eff

• Intent Laundering: AI Safety Datasets Are Not What They Seem

  Shahriar Golchin, Marc Wetter · Feb 17, 2026

  Red Team

  We systematically evaluate the quality of widely used AI safety datasets from two perspectives: in isolation and in practice.

• Enhancing Building Semantics Preservation in AI Model Training with Large Language Model Encodings

  Suhyung Jang, Ghang Lee, Jaekun Lee, Hyunjun Lee · Feb 17, 2026

  Accurate representation of building semantics, encompassing both generic object types and specific subtypes, is essential for effective AI model training in the architecture, engineering, construction, and operation (AECO) industry.

• *-PLUIE: Personalisable metric with Llm Used for Improved Evaluation

  Quentin Lemesle, Léane Jourdan, Daisy Munson, Pierre Alain, Jonathan Chevelu · Feb 17, 2026

  Evaluating the quality of automatically generated text often relies on LLM-as-a-judge (LLM-judge) methods.

• ViTaB-A: Evaluating Multimodal Large Language Models on Visual Table Attribution

  Yahia Alqurnawi, Preetom Biswas, Anmol Rao, Tejas Anvekar, Chitta Baral · Feb 17, 2026

  Multimodal Large Language Models (mLLMs) are often used to answer questions in structured data such as tables in Markdown, JSON, and images.

• GLM-5: from Vibe Coding to Agentic Engineering

  GLM-5-Team, :, Aohan Zeng, Xin Lv, Zhenyu Hou · Feb 17, 2026

  Long Horizon

  We present GLM-5, a next-generation foundation model designed to transition the paradigm of vibe coding to agentic engineering.

• ChartEditBench: Evaluating Grounded Multi-Turn Chart Editing in Multimodal Language Models

  Manav Nitin Kapadnis, Lawanya Baghel, Atharva Naik, Carolyn Rosé · Feb 17, 2026

  Pairwise Preference

  In practice, users iteratively refine visualizations through multi-turn interactions that require maintaining common ground, tracking prior edits, and adapting to evolving preferences.

• Beyond Binary Classification: Detecting Fine-Grained Sexism in Social Media Videos

  Laura De Grazia, Danae Sánchez Villegas, Desmond Elliott, Mireia Farrús, Mariona Taulé · Feb 17, 2026

  Our findings indicate that multimodal LLMs perform competitively with human annotators in identifying nuanced forms of sexism; however, they struggle to capture co-occurring sexist types when these are conveyed through visual cues.

• Under-resourced studies of under-resourced languages: lemmatization and POS-tagging with LLM annotators for historical Armenian, Georgian, Greek and Syriac

  Chahan Vidal-Gorène, Bastien Kindt, Florian Cafiero · Feb 17, 2026

  Using a novel benchmark comprising aligned training and out-of-domain test corpora, we evaluate the performance of foundation models across lemmatization and POS-tagging, and compare them with PIE, a task-specific RNN baseline.

• Causal Effect Estimation with Latent Textual Treatments

  Omri Feldman, Amar Venugopal, Jann Spiess, Amir Feder · Feb 17, 2026

  Understanding the causal effects of text on downstream outcomes is a central task in many applications.

• Recursive Concept Evolution for Compositional Reasoning in Large Language Models

  Sarim Chaudhry · Feb 17, 2026

  Large language models achieve strong performance on many complex reasoning tasks, yet their accuracy degrades sharply on benchmarks that require compositional reasoning, including ARC-AGI-2, GPQA, MATH, BBH, and HLE.

• Rethinking Metrics for Lexical Semantic Change Detection

  Roksana Goworek, Haim Dubossarsky · Feb 17, 2026

  Pairwise Preference

  Lexical semantic change detection (LSCD) increasingly relies on contextualised language model embeddings, yet most approaches still quantify change using a small set of semantic change metrics, primarily Average Pairwise Distance (APD) and

• Proactive Conversational Assistant for a Procedural Manual Task based on Audio and IMU

  Rehana Mahfuz, Yinyi Guo, Erik Visser, Phanidhar Chinchili · Feb 17, 2026

  Real-time conversational assistants for procedural tasks often depend on video input, which can be computationally expensive and compromise user privacy.

• A Content-Based Framework for Cybersecurity Refusal Decisions in Large Language Models

  Noa Linder, Meirav Segal, Omer Antverg, Gil Gekker, Tomer Fichman · Feb 17, 2026

  Large language models and LLM-based agents are increasingly used for cybersecurity tasks that are inherently dual-use.

• Revisiting Northrop Frye's Four Myths Theory with Large Language Models

  Edirlei Soares de Lima, Marco A. Casanova, Antonio L. Furtado · Feb 17, 2026

  Northrop Frye's theory of four fundamental narrative genres (comedy, romance, tragedy, satire) has profoundly influenced literary criticism, yet computational approaches to his framework have focused primarily on narrative patterns rather t

• LLM-to-Speech: A Synthetic Data Pipeline for Training Dialectal Text-to-Speech Models

  Ahmed Khaled Khamis, Hesham Ali · Feb 17, 2026

  Despite the advances in neural text to speech (TTS), many Arabic dialectal varieties remain marginally addressed, with most resources concentrated on Modern Spoken Arabic (MSA) and Gulf dialects, leaving Egyptian Arabic -- the most widely u

• STAPO: Stabilizing Reinforcement Learning for LLMs by Silencing Rare Spurious Tokens

  Shiqi Liu, Zeyu He, Guojian Zhan, Letian Tao, Zhilong Zheng · Feb 17, 2026

  Across six mathematical reasoning benchmarks using Qwen 1.7B, 8B, and 14B base models, STAPO consistently demonstrates superior entropy stability and achieves an average performance improvement of 7.13% ($ρ_{\mathrm{T}}$=1.0, top-p=1.0) and

• Clinically Inspired Symptom-Guided Depression Detection from Emotion-Aware Speech Representations

  Chaithra Nerella, Chiranjeevi Yarra · Feb 17, 2026

  Depression manifests through a diverse set of symptoms such as sleep disturbance, loss of interest, and concentration difficulties.

• Beyond Static Pipelines: Learning Dynamic Workflows for Text-to-SQL

  Yihan Wang, Peiyu Liu, Runyu Chen, Wei Xu · Feb 17, 2026

  Experiments on widely-used Text-to-SQL benchmarks demonstrate that dynamic workflow construction consistently outperforms the best static workflow methods, with especially pronounced gains on complex and out-of-distribution queries.

• RUVA: Personalized Transparent On-Device Graph Reasoning

  Gabriele Conte, Alessio Mattiace, Gianni Carmosino, Potito Aghilar, Giovanni Servedio · Feb 17, 2026

  We propose Ruva, the first "Glass Box" architecture designed for Human-in-the-Loop Memory Curation.

• jina-embeddings-v5-text: Task-Targeted Embedding Distillation

  Mohammad Kalim Akram, Saba Sturua, Nastia Havriushenko, Quentin Herreros, Michael Günther · Feb 17, 2026

  Benchmark scores for the resulting models, jina-embeddings-v5-text-small and jina-embeddings-v5-text-nano, exceed or match the state-of-the-art for models of similar size.

• Perspectives - Interactive Document Clustering in the Discourse Analysis Tool Suite

  Tim Fischer, Chris Biemann · Feb 17, 2026

  Demonstrations

  This paper introduces Perspectives, an interactive extension of the Discourse Analysis Tool Suite designed to empower Digital Humanities (DH) scholars to explore and organize large, unstructured document collections.

• ZeroSyl: Simple Zero-Resource Syllable Tokenization for Spoken Language Modeling

  Nicol Visser, Simon Malan, Danel Slabbert, Herman Kamper · Feb 17, 2026

  ZeroSyl outperforms prior syllabic tokenizers across lexical, syntactic, and narrative benchmarks.

• ExpertWeaver: Unlocking the Inherent MoE in Dense LLMs with GLU Activation Patterns

  Ziyu Zhao, Tong Zhu, Zhi Zhang, Tiantian Fan, Jinluan Yang · Feb 17, 2026

  Mixture-of-Experts (MoE) effectively scales model capacity while preserving computational efficiency through sparse expert activation.

• DependencyAI: Detecting AI Generated Text through Dependency Parsing

  Sara Ahmed, Tracy Hammond · Feb 17, 2026

  To increase interpretability, we analyze feature importance to reveal syntactic structures that distinguish AI-generated from human-written text.

• Fine-Refine: Iterative Fine-grained Refinement for Mitigating Dialogue Hallucination

  Xiangyan Chen, Yujian Gan, Matthew Purver · Feb 17, 2026

  The tendency for hallucination in current large language models (LLMs) negatively impacts dialogue systems.

• LuxMT Technical Report

  Nils Rehlinger · Feb 17, 2026

  To assess translation performance, we construct a novel benchmark covering LB-FR, LB-EN, and LB-FR using human-translated data from Luci, a tourist magazine about Luxembourg.

• Towards Expectation Detection in Language: A Case Study on Treatment Expectations in Reddit

  Aswathy Velutharambath, Amelie Wührl · Feb 17, 2026

  Patients' expectations towards their treatment have a substantial effect on the treatments' success.

• In Agents We Trust, but Who Do Agents Trust? Latent Source Preferences Steer LLM Generations

  Mohammad Aflah Khan, Mahsa Amani, Soumi Das, Bishwamittra Ghosh, Qinyuan Wu · Feb 17, 2026

  Pairwise Preference

  Agents based on Large Language Models (LLMs) are increasingly being deployed as interfaces to information on online platforms.

• TAROT: Test-driven and Capability-adaptive Curriculum Reinforcement Fine-tuning for Code Generation with Large Language Models

  Chansung Park, Juyong Jiang, Fan Wang, Sayak Paul, Jiasi Shen · Feb 17, 2026

  TAROT systematically constructs, for each problem, a four-tier test suite (basic, intermediate, complex, edge), providing a controlled difficulty landscape for curriculum design and evaluation.

• Measuring Social Integration Through Participation: Categorizing Organizations and Leisure Activities in the Displaced Karelians Interview Archive using LLMs

  Joonatan Laato, Veera Schroderus, Jenna Kanerva, Jenni Kauppi, Virpi Lummaa · Feb 17, 2026

  We annotate a gold-standard set to allow for a reliable evaluation, and then test whether large language models can apply the same schema at scale.

• World-Model-Augmented Web Agents with Action Correction

  Zhouzhou Shen, Xueyu Hu, Xiyun Li, Tianqing Fang, Juncheng Li · Feb 17, 2026

  Multi Agent

  Web agents based on large language models have demonstrated promising capability in automating web tasks.

• The Vision Wormhole: Latent-Space Communication in Heterogeneous Multi-Agent Systems

  Xiaoze Liu, Ruowang Zhang, Weichen Yu, Siheng Xiong, Liu He · Feb 17, 2026

  Pairwise Preference Multi Agent

  Multi-Agent Systems (MAS) powered by Large Language Models have unlocked advanced collaborative reasoning, yet they remain shackled by the inefficiency of discrete text communication, which imposes significant runtime overhead and informati

• Making Large Language Models Speak Tulu: Structured Prompting for an Extremely Low-Resource Language

  Prathamesh Devadiga, Paras Chopra · Feb 17, 2026

  Can large language models converse in languages virtually absent from their training data?

• Orchestration-Free Customer Service Automation: A Privacy-Preserving and Flowchart-Guided Framework

  Mengze Hong, Chen Jason Zhang, Zichang Guo, Hanlin Gu, Di Jiang · Feb 17, 2026

  Demonstrations

  Existing approaches either rely on modular system designs with extensive agent orchestration or employ over-simplified instruction schemas, providing limited guidance and poor generalizability.

• Far Out: Evaluating Language Models on Slang in Australian and Indian English

  Deniz Kaya Dilsiz, Dipankar Srirag, Aditya Joshi · Feb 17, 2026

  We present a comprehensive evaluation of slang awareness in Indian English (en-IN) and Australian English (en-AU) across seven state-of-the-art language models.

• NeuroSymActive: Differentiable Neural-Symbolic Reasoning with Active Exploration for Knowledge Graph Question Answering

  Rong Fu, Yang Li, Zeyu Zhang, Jiekai Wu, Yaohua Liu · Feb 17, 2026

  Empirical results on standard KGQA benchmarks show that NeuroSymActive attains strong answer accuracy while reducing the number of expensive graph lookups and model calls compared to common retrieval-augmented baselines.

• Discovering Implicit Large Language Model Alignment Objectives

  Edward Chen, Sanmi Koyejo, Carlos Guestrin · Feb 17, 2026

  Rubric Rating

  To address these limitations, we introduce Obj-Disco, a framework that automatically decomposes an alignment reward signal into a sparse, weighted combination of human-interpretable natural language objectives.

• Prescriptive Scaling Reveals the Evolution of Language Model Capabilities

  Hanlin Zhang, Jikai Jin, Vasilis Syrgkanis, Sham Kakade · Feb 17, 2026

  Using large scale observational evaluations with 5k observational and 2k newly sampled data on model performance, we estimate capability boundaries, high conditional quantiles of benchmark scores as a function of log pre training FLOPs, via

• Mnemis: Dual-Route Retrieval on Hierarchical Graphs for Long-Term LLM Memory

  Zihao Tang, Xin Yu, Ziyu Xiao, Zengxuan Wen, Zelin Li · Feb 17, 2026

  Mnemis achieves state-of-the-art performance across all compared methods on long-term memory benchmarks, scoring 93.9 on LoCoMo and 91.6 on LongMemEval-S using GPT-4.1-mini.

• Extracting Consumer Insight from Text: A Large Language Model Approach to Emotion and Evaluation Measurement

  Stephan Ludwig, Peter J. Danaher, Xiaohao Yang, Yu-Ting Lin, Ehsan Abedin · Feb 17, 2026

  Accurately measuring consumer emotions and evaluations from unstructured text remains a core challenge for marketing research and practice.

• The Information Geometry of Softmax: Probing and Steering

  Kiho Park, Todd Nief, Yo Joong Choe, Victor Veitch · Feb 17, 2026

  This paper concerns the question of how AI systems encode semantic structure into the geometric structure of their representation spaces.

• FrameRef: A Framing Dataset and Simulation Testbed for Modeling Bounded Rational Information Health

  Victor De Lima, Jiqun Liu, Grace Hui Yang · Feb 17, 2026

  Long Horizon

  Within this framework, we construct framing-sensitive agent personas by fine-tuning language models with framing-conditioned loss attenuation, inducing targeted biases while preserving overall task competence.

Recent Daily Archives