Lee

Murmur is an efficient inference system for long-form ASR that resolves the accuracy-latency trade-off by optimizing both inter-chunk processing and intra-chunk attention mechanisms.

The paper introduces a new benchmark for multi-target cross-lingual summarization (MTXLS) and proposes an activation steering method that significantly improves LLM performance by guiding the generation process using English representations.

The paper proposes InSemRAG, an enhanced RAG framework that improves retrieval accuracy and knowledge integrity by incorporating intent-aware retrieval and semantics-preserving chunking, achieving state-of-the-art performance with reduced latency.

The paper introduces ANCHOR, a schema-agnostic system that constructs knowledge graphs from Cyber Threat Intelligence by dynamically discovering and validating against large ontologies, overcoming limitations of fixed schemas and privacy concerns.

The paper reframes Parameter-Efficient Fine-Tuning (PEFT) from a mere cost-saving alternative to a robust architecture for creating persistent, personalized models that layer specific behaviors onto large shared foundation models.

The paper argues that large activation spikes in LLMs are structural vector biases, and proposes a novel quantization framework (INSERTQUANT) to eliminate these spikes, enabling robust low-bit quantization.

The paper addresses Perceptual Judgment Bias in multimodal LLM judges by introducing a new dataset and a unified training framework that forces models to prioritize visual evidence over plausible textual narratives.

The paper proposes Luar, a framework that trains reasoning language models to selectively use English translation only when their direct understanding of a non-English input is unreliable, significantly improving multilingual reasoning.

MORPHOS is a novel autoregressive framework that generates dynamic 3D assets (like meshes and radiance fields) from videos by using a unified 4D representation to ensure temporal consistency and handle changing object topologies.

The paper proposes COVRAG, a depth-based memory retrieval framework that maximizes the coverage of target-view regions to significantly improve long-term geometric consistency in autoregressive long video generation.

The paper introduces CEON, a Circular Economy Ontology Network, designed to improve semantic interoperability and knowledge representation across diverse industry sectors throughout the product life cycle.

The paper introduces AGENTREDBENCH, a dynamic redteaming benchmark that significantly measures indirect prompt injection threats in LLM agents using SaaS integrations, and releases AGENTREDGUARD, a superior defense model.

The paper introduces K-BrowseComp, a new web-browsing agent benchmark of 400 problems grounded in Korean contexts, demonstrating that current frontier LLMs struggle significantly with complex, context-specific web browsing tasks.

The paper introduces AGENTREDBENCH, a dynamic redteaming benchmark that significantly measures indirect prompt injection threats in LLM agents using third-party integrations, and releases AGENTREDGUARD, a superior defense model.

SimuScene introduces a novel compositional 3D reconstruction pipeline that integrates physics simulation directly into the shape and layout estimation process to generate stable, simulation-ready 3D scenes from a single image.

This paper analyzes the conditions under which Bitcoin's security might fail due to miners deviating from honest mining when block rewards decline to zero, concluding that protocol mechanisms can mitigate this risk.

The paper identifies Mid-Session Tool Injection (MSTI) as a novel threat in the WebMCP protocol, demonstrating that attackers can manipulate the visible or perceived set of tools available to AI agents during an active session.

Membrane introduces a self-evolving guardrail using Contrastive Safety Memory (CSM) that generalizes across topical jailbreak variants, achieving superior safety performance while minimizing benign refusal rates.

The paper introduces SlotGCG, an improved jailbreak attack method that systematically searches for the most vulnerable token insertion positions (slots) within a prompt, significantly boosting attack success rates compared to fixed-position methods.

This paper proposes SpatialClaw, a training-free framework for spatial reasoning that enables open-ended, complex 3D/4D spatial reasoning.