SAFECHAIN: Safety of Language Models with Long Chain-of-Thought Reasoning Capabilities
【DL輪読会】 SAFECHAIN: Safety of Language Models with Long Chain-of-Thought Reasoning Cap more
カテゴリー: dls-2025
Prompt-to-SQL Injections in LLM-Integrated Web Applications: Risks and Defenses
【DL輪読会】Prompt-to-SQL Injections in LLM-Integrated Web Applications: Risks and Defense more
Emergent Misalignment: Narrow finetuning can produce broadly misaligned LLMs
【DL輪読会】Emergent Misalignment: Narrow finetuning can produce broadly misaligned LLMs b more
Open-World Reinforcement Learning over Long Short-Term Imagination
【DL輪読会】Open-World Reinforcement Learning over Long Short-Term Imagination by @DeepLea more
Learning Transformer-based World Models with Contrastive Predictive Coding
【DL輪読会】Learning Transformer-based World Models with Contrastive Predictive Coding by more
Titans: Learning to Memorize at Test Time
【DL輪読会】 Titans: Learning to Memorize at Test Time by @DeepLearning2023
Towards Understanding How Knowledge Evolves in Large Vision-Language Models
【DL輪読会】Towards Understanding How Knowledge Evolves in Large Vision-Language Models by more
Language Models Are Capable of Metacognitive Monitoring and Control of Their Internal Activations
【DL輪読会】 Language Models Are Capable of Metacognitive Monitoring and Control of Their more
A continuous time framework for discrete denoising models
【Diffusion勉強会】A continuous time framework for discrete denoising models by @DeepLearn more
From task structures to world models: what do LLMs know?
【DL輪読会】From task structures to world models: what do LLMs know? by @DeepLearning2023