This work presents the first systematic analysis of evaluation pitfalls in multimedia event extraction, identifying three major sources of issues: inconsistent data processing, inconsistent task assumptions, and overly relaxed evaluation settings.
This study reproduces the results of AlphaEdit, a null-space constrained projection method for knowledge editing in language models, and extends the evaluation to newer architectures and longer sequential editing horizons. The authors confirm that AlphaEdit performs as reported within its original scope but identify significant limitations regarding generalization and scalability.
Researchers propose AIGP, a framework using Large Language Models to address interpretability and long-term objective misalignment in e-commerce dynamic pricing. The system employs supervised fine-tuning and a Long-Term Value Estimator trained via offline reinforcement learning to align pricing decisions with business goals.
The authors propose OPID, a framework that extracts skill supervision directly from completed on-policy trajectories to address the sparse reward problem in outcome-based reinforcement learning. By representing trajectory hindsight as hierarchical skills, OPID provides dense, distribution-matched token-level supervision without relying on external memory.
A computational corpus study analyzes vocabulary relationships across eight layers of Bengali and Sanskrit devotional literature from the 8th to 19th centuries, quantifying the historical claim that Buddhist Vajrayana vocabulary was absorbed into the Shakta Tantra tradition. Using TF-IDF character n-gram vectorization on 75 texts, the research provides the first quantitative corroboration of this lexical transmission chain.
The authors propose KARLA, a method enabling large language models to automatically retrieve factual knowledge from an external knowledge base during token generation. This approach allows factual updates without retraining the model and ensures that outputs are traceable to the source data.
This paper details FBK's submission to the IWSLT 2026 Instruction Following shared task, presenting SpeechLLMs designed for both short-form and long-form speech instruction following under constrained settings.
AgentX is a production-deployed multi-agent system designed to automate the iteration of industrial recommender systems, addressing the bottleneck where innovation currently scales linearly with human headcount.
This article introduces a cascaded multi-granularity pruning framework designed to deploy large language models on Industrial Internet of Things (IIoT) edge devices by removing layers, attention heads, and feed-forward channels in a coarse-to-fine order. The method utilizes lightweight low-rank recovery between stages to re-estimate component importance, addressing the collapse of existing structured pruning methods at high compression ratios.
Researchers introduce InfoKV, an entropy-aware framework that compresses key-value caches by combining token-level predictive uncertainty with attention scores to improve long-context reasoning.
This study demonstrates that frozen language models can serve as effective neural predictors for brain activity during natural speech and text comprehension, while distinguishing predictive utility from claims about shared neural organization. The analysis of MEG and ECoG data revealed widespread positive prediction gains over low-level baselines, though participant-level advantages were localized rather than uniform.
This study audits the reliability of eight state-of-the-art Automatic Speech Recognition models on real-world psychiatric interview data in Kannada, Hindi, and Indian English. The results reveal substantial variability across models and languages, with some systems performing competitively in Indian English but failing in regional speech.
Vision-language models frequently generate hallucinated outputs where text and images are misaligned, necessitating methods that not only detect these errors but also explain them and localize visual evidence. The authors introduce GAVEL, a task designed to jointly address verification, explanation, and localization for image-text pairs, accompanied by a corresponding dataset and benchmark.
This study investigates whether non-expert malicious actors can successfully jailbreak large language models by using bandit algorithms to select optimal attacks and enhance queries. The authors propose a novel attack strategy based on the multi-armed bandit framework to efficiently learn the best jailbreak from a large choice set through noisy exploration.
Researchers propose a term-centric framework for inducing hierarchical taxonomies from diverse text sources, addressing the limitations of existing methods that rely on document-level representations. This approach maps documents into a shared representation space via automatic term extraction to enable robust cross-source alignment and construct interpretable hierarchies.
A new study reveals significant safety and fairness gaps in multilingual speech models, finding that only 8% of state-of-the-art releases document any multilingual analysis. To address this, the authors introduce RedVox, a benchmark built on real voices covering unsafe requests across five languages.
The article introduces Einstein World Models (EWMs), a framework designed to enhance large language model reasoning by integrating visual-temporal rollouts into the reasoning trace. This approach allows models to utilize visual thought experiments as inspectable hypotheses to complement text-based processing.
This study investigates how semantically similar concerns presented through different contextual framings elicit varying responses from instruction-tuned large language models, potentially challenging system reliability. Using controlled matched prompts and layer-wise probing analyses, the authors demonstrate that framing systematically alters interpretive response tendencies across multiple model architectures.
Researchers propose ReaORE, a framework for open relation extraction that utilizes large reasoning models to achieve reliable generalization to unseen relation types. The method addresses limitations of current clustering and direct generation approaches through a coarse-to-fine reasoning process.
This study investigates the presence and structure of emotion vectors in open-weight large language models, specifically Apertus-8B-Instruct-2509 and Gemma-4-E4B-it. The research confirms that these models encode valence geometry with high correlation to human psychological structures, approaching the levels previously observed in Claude Sonnet 4.5.