2025-03-07s
2025
Lost in Literalism: How Supervised Training Shapes Translationese in LLMs
·3432 words·17 mins·
loading
·
loading
AI Generated
🤗 Daily Papers
Natural Language Processing
Machine Translation
🏢 Shanghai AI Laboratory
LLMs show translationese due to supervised training biases. Polishing references and filtering unnatural instances can mitigate this issue.
IFIR: A Comprehensive Benchmark for Evaluating Instruction-Following in Expert-Domain Information Retrieval
·5266 words·25 mins·
loading
·
loading
AI Generated
🤗 Daily Papers
Natural Language Processing
Information Extraction
🏢 School of Advanced Interdisciplinary Sciences, University of Chinese Academy of Sciences
IFIR: a new benchmark for instruction-following retrieval in expert domains, revealing current model limitations.
FuseChat-3.0: Preference Optimization Meets Heterogeneous Model Fusion
·449 words·3 mins·
loading
·
loading
AI Generated
🤗 Daily Papers
Natural Language Processing
Large Language Models
🏢 School of Computer Science and Engineering, Sun Yat-Sen University, China
FuseChat-3.0: Heterogeneous model fusion boosts LLM performance via preference optimization, creating efficient and powerful language models.
EgoLife: Towards Egocentric Life Assistant
·3562 words·17 mins·
loading
·
loading
AI Generated
🤗 Daily Papers
Multimodal Learning
Human-AI Interaction
🏢 NTU S-Lab
EgoLife: Ultra-long egocentric dataset & benchmark enabling AI assistants to understand and enhance daily life. Datasets and models released!
LINGOLY-TOO: Disentangling Memorisation from Reasoning with Linguistic Templatisation and Orthographic Obfuscation
·4618 words·22 mins·
loading
·
loading
AI Generated
🤗 Daily Papers
Natural Language Processing
Large Language Models
🏢 University of Oxford
LINGOLY-TOO: A new benchmark to disentangle memorization from reasoning in LLMs using linguistic templatization and orthographic obfuscation.
Identifying Sensitive Weights via Post-quantization Integral
·2603 words·13 mins·
loading
·
loading
AI Generated
🤗 Daily Papers
Machine Learning
Deep Learning
🏢 Tsinghua University
PQI: Accurately identify sensitive weights in post-quantization to enhance LLM compression & performance!