Mosaic Pruning: A Hierarchical Framework for Generalizable Pruning of Mixture-of-Experts Models

A recent study has introduced an automated pipeline for generating large-scale, psychologically-grounded multi-turn jailbreak datasets for Large Language Models (LLMs). This approach leverages psychological principles like Foot-in-the-Door (FITD) to create a benchmark of 1,500 scenarios, revealing significant vulnerabilities in models, particularly those in the GPT family, when subjected to multi-turn conversational attacks.

Recent research has highlighted significant vulnerabilities in Large Language Models (LLMs), particularly concerning prompt injection and jailbreaking attacks. This review categorizes various attack methods and evaluates defense strategies, including prompt filtering and self-regulation, to mitigate these risks.

The introduction of HERMES, a Heterogeneous Multi-stage LLM inference Execution Simulator, marks a significant advancement in optimizing inference pipelines for Large Language Models (LLMs). This tool addresses the limitations of existing simulators by accurately modeling diverse request stages, including Retrieval Augmented Generation (RAG) and key-value cache retrieval, across complex hardware architectures.

A new framework called PersonaPulse has been introduced to optimize prompts for Large Language Models (LLMs), enhancing their ability to express realistic personality traits. This approach iteratively refines role-play prompts while using a situational response benchmark for evaluation, demonstrating improved performance over previous methods based on psychological personality descriptions.

A new framework called Mixture of Attention Spans (MoA) has been proposed to enhance the efficiency of Large Language Models (LLMs) by optimizing inference through heterogeneous sliding-window lengths. This approach addresses the limitations of existing methods that use a uniform window length, which fails to capture the diverse attention patterns in LLMs, particularly in long-context scenarios.

A systematic study has been conducted on compression ordering for large language models (LLMs), specifically focusing on the Qwen2.5 3B model. The research evaluates various compression techniques such as knowledge distillation, structured pruning, and low-bit quantization, analyzing their performance both independently and in combination. The findings indicate that quantization offers the highest standalone compression, while the sequence of techniques significantly impacts the final model quality.

A recent study published on arXiv investigates prompt fairness in Large Language Models (LLMs), revealing significant disparities in response quality based on how prompts are phrased by different users. The research employs information-theoretic metrics to assess subgroup sensitivity and cross-group consistency, highlighting structural inequities in model behavior across various demographic subgroups.

Large Language Models (LLMs) exhibit strong generalization across various tasks, yet their internal decision-making processes remain unclear. A recent study investigates the geometry of hidden representations in LLMs, focusing on intrinsic dimension (ID) in multiple-choice question answering (MCQA) settings. The research reveals a consistent ID pattern across different transformer models, indicating how LLMs project linguistic inputs onto structured, low-dimensional manifolds.