Text understanding and language tasks
![[2602.00750] Bypassing Prompt Injection Detectors through Evasive Injections](https://arxiv.org/static/browse/0.3.4/images/arxiv-logo-fb.png){kind=logo}
Abstract page for arXiv paper 2602.00750: Bypassing Prompt Injection Detectors through Evasive Injections
Abstract page for arXiv paper 2512.18640: Geometric-Photometric Event-based 3D Gaussian Ray Tracing
Abstract page for arXiv paper 2511.08225: Benchmarking Educational LLMs with Analytics: A Case Study on Gender Bias in Feedback
The paper presents MomentMix, a novel augmentation technique using Length-Aware DETR to enhance video moment retrieval, particularly for ...
The paper presents VALTEST, a framework for validating test cases generated by large language models (LLMs) using semantic entropy, impro...
The G-reasoner paper introduces a unified framework that enhances reasoning over graph-structured knowledge using a new graph foundation ...
The paper introduces Versor, a novel geometric sequence architecture that leverages Conformal Geometric Algebra for enhanced performance ...
The paper explores the paradox of scaling large language models (LLMs) in context compression, revealing that larger models may reduce th...
The paper presents LiveMCPBench, a benchmark designed to evaluate the capabilities of agents using Model Context Protocol (MCP) tools in ...
The paper presents FHIR-RAG-MEDS, a system that integrates HL7 FHIR with Retrieval-Augmented Generation models to enhance personalized me...
The paper presents a novel post-training method that enhances transformer attention sparsity while maintaining performance, revealing ins...
The paper presents a novel method for learning and sampling from probability distributions on the simplex, utilizing smooth bijections to...
The paper explores a novel approach to learning answer generation from correct demonstrations, formalizing it as imitation learning withi...
This article presents a novel approach to improving masked diffusion models (MDMs) for language modeling by introducing a learned schedul...
This article explores the statistical advantages of softmax attention mechanisms in large language models, particularly in single-locatio...
The paper presents SPARTA, a novel framework for generating scalable benchmarks for tree-structured multi-hop question answering (QA) ove...
This article presents rBridge, a small proxy model that predicts reasoning performance in large language models (LLMs), demonstrating sig...
The paper presents MovieTeller, a novel framework for generating movie synopses using tool-augmented progressive abstraction to enhance c...
This paper investigates why Diffusion Language Models (DLMs) often default to autoregressive decoding instead of utilizing their potentia...
This article explores RL-Obfuscation, a method for training language models to evade latent-space monitors that detect undesirable behavi...
This study examines the relationship between fluency and accuracy in L2 Mandarin prosody, revealing that while learners may achieve quant...
This paper presents a novel approach to long-form Bengali Automatic Speech Recognition (ASR) and speaker diarization, introducing a compr...
The paper introduces Affine-Scaled Attention, a novel approach to Transformer attention that enhances flexibility and stability by modify...
Get the latest news, tools, and insights delivered to your inbox.
Daily or weekly digest • Unsubscribe anytime