KI-Forschungspapiere Täglich

FuseChat-3.0: Preference Optimization Meets Heterogeneous Model Fusion

LLMVoX: Autoregressive Streaming Text-to-Speech Model for Any LLM

Audio Flamingo 2: An Audio-Language Model with Long-Audio Understanding and Expert Reasoning Abilities

The Best of Both Worlds: Integrating Language Models and Diffusion Models for Video Generation

HybridNorm: Towards Stable and Efficient Transformer Training via Hybrid Normalization

Dedicated Feedback and Edit Models Empower Inference-Time Scaling for Open-Ended General-Domain Tasks

LINGOLY-TOO: Disentangling Memorisation from Reasoning with Linguistic Templatisation and Orthographic Obfuscation

PokéChamp: an Expert-level Minimax Language Agent

IFIR: A Comprehensive Benchmark for Evaluating Instruction-Following in Expert-Domain Information Retrieval

Identifying Sensitive Weights via Post-quantization Integral

L^2M: Mutual Information Scaling Law for Long-Context Language Modeling

LLM as a Broken Telephone: Iterative Generation Distorts Information

On the Acquisition of Shared Grammatical Representations in Bilingual Language Models

Token-Efficient Long Video Understanding for Multimodal LLMs

How to Steer LLM Latents for Hallucination Detection?

Union of Experts: Adapting Hierarchical Routing to Equivalently Decomposed Transformer

EgoLife: Towards Egocentric Life Assistant

START: Self-taught Reasoner with Tools

Understanding and Predicting Derailment in Toxic Conversations on GitHub

Combining Flow Matching and Transformers for Efficient Solution of Bayesian Inverse Problems

Lost in Literalism: How Supervised Training Shapes Translationese in LLMs