Large Language Model
UniBench: VisualReasoningRequiresRethinking Vision-LanguageBeyondScaling
Wefind that while scaling training data ormodel size can boost many vision-language model capabilities, scaling offers little benefit for reasoning or relations. Surprisingly, we also discover today's best VLMs struggle on simple digit recognition and counting tasks, e.g. MNIST, which much simpler networks can solve.
On Optimal Caching and Model Multiplexing for Large Model Inference
By combining a caching algorithm, namely Greedy Dual Size with Frequency (GDSF) or Least Expected Cost (LEC), with a model multiplexer, we achieve optimal rates in both offline and online settings.
SlowFocus: Enhancing Fine-grained Temporal Understanding in Video LLM
However, current Vid-LLMs struggle to simultaneously retain high-quality frame-level semantic information ( i.e., a sufficient
Online Iterative Reinforcement Learning from Human Feedback with General Preference Model
The second step is reward modeling, which is the origin of the name "reward-based".
Mitigating Reward Overoptimization via Lightweight Uncertainty Estimation
Reinforcement Learning from Human Feedback (RLHF) has been pivotal in aligning Large Language Models with human values but often suffers from overopti-mization due to its reliance on a proxy reward model.
