HFEPX Daily Archive: 2026-02-20

Research Utility Snapshot

Papers Published On This Date

• PolyFrame at MWE-2026 AdMIRe 2: When Words Are Not Enough: Multimodal Idiom Disambiguation

  Nina Hosseini-Kivanani · Feb 20, 2026

  Multimodal models struggle with idiomatic expressions due to their non-compositional meanings, a challenge amplified in multilingual settings.

• DP-RFT: Learning to Generate Synthetic Text via Differentially Private Reinforcement Fine-Tuning

  Fangyuan Xu, Sihao Chen, Zinan Lin, Taiwei Shi, Sydney Graham · Feb 20, 2026

  Differentially private (DP) synthetic data generation plays a pivotal role in developing large language models (LLMs) on private data, where data owners cannot provide eyes-on access to individual examples.

• Diagnosing LLM Reranker Behavior Under Fixed Evidence Pools

  Baris Arat, Emre Sefer · Feb 20, 2026

  Standard reranking evaluations study how a reranker orders candidates returned by an upstream retriever.

• Luna-2: Scalable Single-Token Evaluation with Small Language Models

  Vatsal Goel, Rishon Dsouza, Nikhil Ega, Amey Ramesh Rambatla, Rob Friel · Feb 20, 2026

  Real-time guardrails require evaluation that is accurate, cheap, and fast - yet today's default, LLM-as-a-judge (LLMAJ), is slow, expensive, and operationally non-deterministic due to multi-token generation.

• Hierarchical Reward Design from Language: Enhancing Alignment of Agent Behavior with Human Specifications

  Zhiqin Qian, Ryan Diaz, Sangwon Seo, Vaibhav Unhelkar · Feb 20, 2026

  Pairwise Preference Long Horizon

  When training artificial intelligence (AI) to perform tasks, humans often care not only about whether a task is completed but also how it is performed.

• VIRAASAT: Traversing Novel Paths for Indian Cultural Reasoning

  Harshul Raj Surana, Arijit Maji, Aryan Vats, Akash Ghosh, Sriparna Saha · Feb 20, 2026

  Existing Cultural benchmarks are (i) Manually crafted, (ii) contain single-hop questions testing factual recall, and (iii) prohibitively costly to scale, leaving this deficiency largely unmeasured.

• RVR: Retrieve-Verify-Retrieve for Comprehensive Question Answering

  Deniz Qian, Hung-Ting Chen, Eunsol Choi · Feb 20, 2026

  Our method outperforms baselines, including agentic search approaches, achieving at least 10% relative and 3% absolute gain in complete recall percentage on a multi-answer retrieval dataset (QAMPARI).

• SPQ: An Ensemble Technique for Large Language Model Compression

  Jiamin Yao, Eren Gultepe · Feb 20, 2026

  Applied to LLaMA-2-7B, SPQ achieves up to 75% memory reduction while maintaining or improving perplexity (e.g., WikiText-2 5.47 to 4.91) and preserving accuracy on downstream benchmarks such as C4, TruthfulQA, and GSM8K.

• Subgroups of $U(d)$ Induce Natural RNN and Transformer Architectures

  Joshua Nunley · Feb 20, 2026

  This paper presents a direct framework for sequence models with hidden states on closed subgroups of U(d).

• Validating Political Position Predictions of Arguments

  Jordan Robinson, Angus R. Williams, Katie Atkinson, Anthony G. Cohn · Feb 20, 2026

  Pairwise Preference

  Real-world knowledge representation often requires capturing subjective, continuous attributes -- such as political positions -- that conflict with pairwise validation, the widely accepted gold standard for human evaluation.

• Vichara: Appellate Judgment Prediction and Explanation for the Indian Judicial System

  Pavithra PM Nair, Preethu Rose Anish · Feb 20, 2026

  Vichara surpasses existing judgment prediction benchmarks on both datasets, with GPT-4o mini achieving the highest performance (F1: 81.5 on PredEx, 80.3 on ILDC_expert), followed by Llama-3.1-8B.

• On the "Induction Bias" in Sequence Models

  M. Reza Ebrahimi, Michaël Defferrard, Sunny Panchal, Roland Memisevic · Feb 20, 2026

  Together, these results demonstrate that state tracking remains a fundamental challenge for transformers, even when training and evaluation distributions match.

• Predicting Contextual Informativeness for Vocabulary Learning using Deep Learning

  Tao Wu, Adam Kapelner · Feb 20, 2026

  In summary, we demonstrate that a modern embedding model on neural network architecture, when guided by human supervision, results in a low-cost large supply of near-perfect contexts for teaching vocabulary for a variety of target words.

• PsihoRo: Depression and Anxiety Romanian Text Corpus

  Alexandra Ciobotaru, Ana-Maria Bucur, Liviu P. Dinu · Feb 20, 2026

  Psychological corpora in NLP are collections of texts used to analyze human psychology, emotions, and mental health.

• VeriSoftBench: Repository-Scale Formal Verification Benchmarks for Lean

  Yutong Xin, Qiaochu Chen, Greg Durrett, Işil Dillig · Feb 20, 2026

  However, most benchmarks for LLM-based proof automation are drawn from mathematics in the Mathlib ecosystem, whereas proofs in software verification are developed inside definition-rich codebases with substantial project-specific libraries.

• On the Semantic and Syntactic Information Encoded in Proto-Tokens for One-Step Text Reconstruction

  Ivan Bondarenko, Egor Palkin, Fedor Tikunov · Feb 20, 2026

  Autoregressive large language models (LLMs) generate text token-by-token, requiring n forward passes to produce a sequence of length n.

• Analyzing and Improving Chain-of-Thought Monitorability Through Information Theory

  Usman Anwar, Tim Bakker, Dana Kianfar, Cristina Pinneri, Christos Louizos · Feb 20, 2026

  Chain-of-thought (CoT) monitors are LLM-based systems that analyze reasoning traces to detect when outputs may exhibit attributes of interest, such as test-hacking behavior during code generation.

• Context-Aware Mapping of 2D Drawing Annotations to 3D CAD Features Using LLM-Assisted Reasoning for Manufacturing Automation

  Muhammad Tayyab Khan, Lequn Chen, Wenhe Feng, Seung Ki Moon · Feb 20, 2026

  When deterministic scoring cannot resolve an ambiguity, the system escalates to multimodal and constrained large-language-model reasoning, followed by a single human-in-the-loop (HITL) review step.

• Simplifying Outcomes of Language Model Component Analyses with ELIA

  Aaron Louis Eidt, Nils Feldhus · Feb 20, 2026

  Pairwise Preference

  The effectiveness of this approach was empirically validated through a mixed-methods user study, which revealed a clear preference for interactive, explorable interfaces over simpler, static visualizations.

• Thinking by Subtraction: Confidence-Driven Contrastive Decoding for LLM Reasoning

  Lexiang Tang, Weihao Gao, Bingchen Zhao, Lu Ma, Qiao jin · Feb 20, 2026

  Experiments show that CCD significantly improves accuracy across mathematical reasoning benchmarks while substantially reducing output length, with minimal KV-cache overhead.

• Information-Theoretic Storage Cost in Sentence Comprehension

  Kohei Kajikawa, Shinnosuke Isono, Ethan Gotlieb Wilcox · Feb 20, 2026

  Real-time sentence comprehension imposes a significant load on working memory, as comprehenders must maintain contextual information to anticipate future input.

• Improving Sampling for Masked Diffusion Models via Information Gain

  Kaisen Yang, Jayden Teoh, Kaicheng Yang, Yitong Zhang, Alex Lamb · Feb 20, 2026

  Extensive evaluations across diverse architectures and tasks (reasoning, coding, creative writing, and image generation) demonstrate that Info-Gain Sampler consistently outperforms existing samplers for MDMs.

• Click it or Leave it: Detecting and Spoiling Clickbait with Informativeness Measures and Large Language Models

  Wojciech Michaluk, Tymoteusz Urban, Mateusz Kubita, Soveatin Kuntur, Anna Wroblewska · Feb 20, 2026

  Clickbait headlines degrade the quality of online information and undermine user trust.

• FENCE: A Financial and Multimodal Jailbreak Detection Dataset

  Mirae Kim, Seonghun Jeong, Youngjun Kwak · Feb 20, 2026

  Red Team

  A baseline detector trained on FENCE achieves 99 percent in-distribution accuracy and maintains strong performance on external benchmarks, underscoring the dataset's robustness for training reliable detection models.

• The Statistical Signature of LLMs

  Ortal Hadad, Edoardo Loru, Jacopo Nudo, Niccolò Di Marco, Matteo Cinelli · Feb 20, 2026

  We analyze compression behavior across three progressively more complex information ecosystems: controlled human-LLM continuations, generative mediation of a knowledge infrastructure (Wikipedia vs.

• Detecting Contextual Hallucinations in LLMs with Frequency-Aware Attention

  Siya Qi, Yudong Chen, Runcong Zhao, Qinglin Zhu, Zhanghao Hu · Feb 20, 2026

  Experiments on the RAGTruth and HalluRAG benchmarks show that our approach achieves performance gains over verification-based, internal-representation-based, and attention-based methods across models and tasks.

• Agentic Adversarial QA for Improving Domain-Specific LLMs

  Vincent Grari, Ciprian Tomoiaga, Sylvain Lamprier, Tatsunori Hashimoto, Marcin Detyniecki · Feb 20, 2026

  Evaluation on specialized subsets of the LegalBench corpus demonstrates that our method achieves greater accuracy with substantially fewer synthetic samples.

• Perceived Political Bias in LLMs Reduces Persuasive Abilities

  Matthew DiGiuseppe, Joshua Robison · Feb 20, 2026

  Conversational AI has been proposed as a scalable way to correct public misconceptions and spread misinformation.

• Gradient Regularization Prevents Reward Hacking in Reinforcement Learning from Human Feedback and Verifiable Rewards

  Johannes Ackermann, Michael Noukhovitch, Takashi Ishida, Masashi Sugiyama · Feb 20, 2026

  Reinforcement Learning from Human Feedback (RLHF) or Verifiable Rewards (RLVR) are two key steps in the post-training of modern Language Models (LMs).

• Towards More Standardized AI Evaluation: From Models to Agents

  Ali El Filali, Inès Bedar · Feb 20, 2026

  Evaluation is no longer a final checkpoint in the machine learning lifecycle.

• NIMMGen: Learning Neural-Integrated Mechanistic Digital Twins with LLMs

  Zihan Guan, Rituparna Datta, Mengxuan Hu, Shunshun Liu, Aiying Zhang · Feb 20, 2026

  Recent work has explored LLM-based agentic frameworks to automatically construct mechanistic models from data; however, existing problem settings substantially oversimplify real-world conditions, leaving it unclear whether LLM-generated mec

• Decomposing Retrieval Failures in RAG for Long-Document Financial Question Answering

  Amine Kobeissi, Philippe Langlais · Feb 20, 2026

  Retrieval-augmented generation is increasingly used for financial question answering over long regulatory filings, yet reliability depends on retrieving the exact context needed to justify answers in high stakes settings.

• CUICurate: A GraphRAG-based Framework for Automated Clinical Concept Curation for NLP applications

  Victoria Blake, Mathew Miller, Jamie Novak, Sze-yuan Ooi, Blanca Gallego · Feb 20, 2026

  Expert Verification

  The framework was evaluated on five lexically heterogeneous clinical concepts against a manually curated benchmark and gold-standard concept sets.

• Analyzing LLM Instruction Optimization for Tabular Fact Verification

  Xiaotang Du, Giwon Hong, Wai-Chung Kwan, Rohit Saxena, Ivan Titov · Feb 20, 2026

  We study three optimizers from the DSPy framework -- COPRO, MiPROv2, and SIMBA -- across four benchmarks and three model families.

• Condition-Gated Reasoning for Context-Dependent Biomedical Question Answering

  Jash Rajesh Parekh, Wonbin Kweon, Joey Chan, Rezarta Islamaj, Robert Leaman · Feb 20, 2026

  Existing benchmarks do not evaluate such conditional reasoning, and retrieval-augmented or graph-based methods lack explicit mechanisms to ensure that retrieved knowledge is applicable to given context.

• Improving Neural Topic Modeling with Semantically-Grounded Soft Label Distributions

  Raymond Li, Amirhossein Abaskohi, Chuyuan Li, Gabriel Murray, Giuseppe Carenini · Feb 20, 2026

  Traditional neural topic models are typically optimized by reconstructing the document's Bag-of-Words (BoW) representations, overlooking contextual information and struggling with data sparsity.

• Games That Teach, Chats That Convince: Comparing Interactive and Static Formats for Persuasive Learning

  Seyed Hossein Alavi, Zining Wang, Shruthi Chockkalingam, Raymond T. Ng, Vered Shwartz · Feb 20, 2026

  Interactive systems such as chatbots and games are increasingly used to persuade and educate on sustainability-related topics, yet it remains unclear how different delivery formats shape learning and persuasive outcomes when content is held

Recent Daily Archives