Vision-language models have achieved remarkable success in multi-modal representation learning from large-scale pairs of visual scenes and linguistic descriptions. However, they still struggle to simultaneously express two distinct types of semantic structures: the hierarchy within a concept family (e.g., dog $\preceq$ mammal $\preceq$ animal) and the compositionality across different concept families (e.g., "a dog in a car" $\preceq$ dog, car). Recent works have addressed this challenge by employing hyperbolic space, which efficiently captures tree-like hierarchy, yet its suitability for representing compositionality remains unclear. To resolve this dilemma, we propose PHyCLIP, which employs an $\ell_1$-Product metric on a Cartesian product of Hyperbolic factors. With our design, intra-family hierarchies emerge within individual hyperbolic factors, and cross-family composition is captured by the $\ell_1$-product metric, analogous to a Boolean algebra. Experiments on zero-shot classification, retrieval, hierarchical classification, and compositional understanding tasks demonstrate that PHyCLIP outperforms existing single-space approaches and offers more interpretable structures in the embedding space.

We introduce and study artificial impressions--patterns in LLMs' internal representations of prompts that resemble human impressions and stereotypes based on language. We fit linear probes on generated prompts to predict impressions according to the two-dimensional Stereotype Content Model (SCM). Using these probes, we study the relationship between impressions and downstream model behavior as well as prompt features that may inform such impressions. We find that LLMs inconsistently report impressions when prompted, but also that impressions are more consistently linearly decodable from their hidden representations. Additionally, we show that artificial impressions of prompts are predictive of the quality and use of hedging in model responses. We also investigate how particular content, stylistic, and dialectal features in prompts impact LLM impressions.

Accurate recognition of biomedical named entities is critical for medical information extraction and knowledge discovery. However, existing methods often struggle with nested entities, entity boundary ambiguity, and cross-lingual generalization. In this paper, we propose a unified Biomedical Named Entity Recognition (BioNER) framework based on Large Language Models (LLMs). We first reformulate BioNER as a text generation task and design a symbolic tagging strategy to jointly handle both flat and nested entities with explicit boundary annotation. To enhance multilingual and multi-task generalization, we perform bilingual joint fine-tuning across multiple Chinese and English datasets. Additionally, we introduce a contrastive learning-based entity selector that filters incorrect or spurious predictions by leveraging boundary-sensitive positive and negative samples. Experimental results on four benchmark datasets and two unseen corpora show that our method achieves state-of-the-art performance and robust zero-shot generalization across languages. The source codes are freely available at https://github.com/dreamer-tx/LLMNER.

The complexity of the Generally Accepted Accounting Principles (GAAP) and the hierarchical structure of eXtensible Business Reporting Language (XBRL) filings make financial auditing increasingly difficult to automate and verify. While large language models (LLMs) have demonstrated strong capabilities in unstructured text understanding, their ability to reason over structured, interdependent, and taxonomy-driven financial documents remains largely unexplored. To fill this gap, we introduce FinAuditing, the first taxonomy-aligned, structure-aware, multi-document benchmark for evaluating LLMs on financial auditing tasks. Built from real US-GAAP-compliant XBRL filings, FinAuditing defines three complementary subtasks, FinSM for semantic consistency, FinRE for relational consistency, and FinMR for numerical consistency, each targeting a distinct aspect of structured auditing reasoning. We further propose a unified evaluation framework integrating retrieval, classification, and reasoning metrics across these subtasks. Extensive zero-shot experiments on 13 state-of-the-art LLMs reveal that current models perform inconsistently across semantic, relational, and mathematical dimensions, with accuracy drops of up to 60-90% when reasoning over hierarchical multi-document structures. Our findings expose the systematic limitations of modern LLMs in taxonomy-grounded financial reasoning and establish FinAuditing as a foundation for developing trustworthy, structure-aware, and regulation-aligned financial intelligence systems. The benchmark dataset is available at Hugging Face.

A big part of the high - level programming repositories used by software developers over the world every day have more than 2000 files; the largest files from those repositories could have 5,000 or more lines of code -- those scales exceed LLM context window size by magnitude of 1,000 - 10,000 times. Based on such limitations, application of highly productive and smart LLM to large code bases becomes a challenge which stays at the front door of automatization of the software development process [8, 3]. Keeping this in mind, we developed a solution that allows us to apply a certain level of automatization and simplification of the development of large code bases for software developers. The most complete and main task that our system is capable of is automatic bug - fix and feature addition / enhancement using only a short user description. The given task can be split into two smaller tasks: (1) retrieval of relevant source code repository files to a Natural Language (NL) user query / issue description; (2) applying changes to a set of files selected in Step #1.

Many important applications across science, data analytics, and AI workloads depend on distributed matrix multiplication. Prior work has developed a large array of algorithms suitable for different problem sizes and partitionings including 1D, 2D, 1.5D, and 2.5D algorithms. A limitation of current work is that existing algorithms are limited to a subset of partitionings. Multiple algorithm implementations are required to support the full space of possible partitionings. If no algorithm implementation is available for a particular set of partitionings, one or more operands must be redistributed, increasing communication costs. This paper presents a universal one-sided algorithm for distributed matrix multiplication that supports all combinations of partitionings and replication factors. Our algorithm uses slicing (index arithmetic) to compute the sets of overlapping tiles that must be multiplied together. This list of local matrix multiplies can then either be executed directly, or reordered and lowered to an optimized IR to maximize overlap. We implement our algorithm using a high-level C++-based PGAS programming framework that performs direct GPU-to-GPU communication using intra-node interconnects. We evaluate performance for a wide variety of partitionings and replication factors, finding that our work is competitive with PyTorch DTensor, a highly optimized distributed tensor library targeting AI models.

Peer review is the cornerstone of scientific publishing, yet it suffers from inconsistencies, reviewer subjectivity, and scalability challenges. We introduce ReviewerToo, a modular framework for studying and deploying AI-assisted peer review to complement human judgment with systematic and consistent assessments. ReviewerToo supports systematic experiments with specialized reviewer personas and structured evaluation criteria, and can be partially or fully integrated into real conference workflows. We validate ReviewerToo on a carefully curated dataset of 1,963 paper submissions from ICLR 2025, where our experiments with the gpt-oss-120b model achieves 81.8% accuracy for the task of categorizing a paper as accept/reject compared to 83.9% for the average human reviewer. Additionally, ReviewerToo-generated reviews are rated as higher quality than the human average by an LLM judge, though still trailing the strongest expert contributions. Our analysis highlights domains where AI reviewers excel (e.g., fact-checking, literature coverage) and where they struggle (e.g., assessing methodological novelty and theoretical contributions), underscoring the continued need for human expertise. Based on these findings, we propose guidelines for integrating AI into peer-review pipelines, showing how AI can enhance consistency, coverage, and fairness while leaving complex evaluative judgments to domain experts. Our work provides a foundation for systematic, hybrid peer-review systems that scale with the growth of scientific publishing.

Large language models (LLMs) remain vulnerable to multi-turn jailbreaking attacks that exploit conversational context to bypass safety constraints gradually. These attacks target different harm categories (like malware generation, harassment, or fraud) through distinct conversational approaches (educational discussions, personal experiences, hypothetical scenarios). Existing multi-turn jailbreaking methods often rely on heuristic or ad hoc exploration strategies, providing limited insight into underlying model weaknesses. The relationship between conversation patterns and model vulnerabilities across harm categories remains poorly understood. We propose Pattern Enhanced Chain of Attack (PE-CoA), a framework of five conversation patterns to construct effective multi-turn jailbreaks through natural dialogue. Evaluating PE-CoA on twelve LLMs spanning ten harm categories, we achieve state-of-the-art performance, uncovering pattern-specific vulnerabilities and LLM behavioral characteristics: models exhibit distinct weakness profiles where robustness to one conversational pattern does not generalize to others, and model families share similar failure modes. These findings highlight limitations of safety training and indicate the need for pattern-aware defenses. Code available on: https://github.com/Ragib-Amin-Nihal/PE-CoA

The ventral, dorsal, and lateral streams in high-level human visual cortex are implicated in distinct functional processes. Yet, deep neural networks (DNNs) trained on a single task model the entire visual system surprisingly well, hinting at common computational principles across these pathways. To explore this inconsistency, we applied a novel sparse decomposition approach to identify the dominant components of visual representations within each stream. Consistent with traditional neuroscience research, we find a clear difference in component response profiles across the three visual streams -- identifying components selective for faces, places, bodies, text, and food in the ventral stream; social interactions, implied motion, and hand actions in the lateral stream; and some less interpretable components in the dorsal stream. Building on this, we introduce Sparse Component Alignment (SCA), a new method for measuring representational alignment between brains and machines that better captures the latent neural tuning of these two visual systems. Using SCA, we find that standard visual DNNs are more aligned with the ventral than either dorsal or lateral representations. SCA reveals these distinctions with greater resolution than conventional population-level geometry, offering a measure of representational alignment that is sensitive to a system's underlying axes of neural tuning.

Understanding the internal representations of large language models is crucial for ensuring their reliability and safety, with sparse autoencoders (SAEs) emerging as a promising interpretability approach. However, current SAE training methods face feature absorption, where features (or neurons) are absorbed into each other to minimize $L_1$ penalty, making it difficult to consistently identify and analyze model behaviors. We introduce Adaptive Temporal Masking (ATM), a novel training approach that dynamically adjusts feature selection by tracking activation magnitudes, frequencies, and reconstruction contributions to compute importance scores that evolve over time. ATM applies a probabilistic masking mechanism based on statistical thresholding of these importance scores, creating a more natural feature selection process. Through extensive experiments on the Gemma-2-2b model, we demonstrate that ATM achieves substantially lower absorption scores compared to existing methods like TopK and JumpReLU SAEs, while maintaining excellent reconstruction quality. These results establish ATM as a principled solution for learning stable, interpretable features in neural networks, providing a foundation for more reliable model analysis.