test-time-computation

🦾 A Dual-System VLA with System2 Thinking

Materials for the course Principles of AI: LLMs at UPenn (Stat 9911, Spring 2025). LLM architectures, training paradigms (pre- and post-training, alignment), test-time computation, reasoning, safety and robustness (jailbreaking, oversight, uncertainty), representations, interpretability (circuits), etc.

This repo is the code for T-SCEND, a novel framework that significantly improves diffusion model’s reasoning capabilities with better energy-based training and scaling up test-time computation.

Accelerate test-time-compute with batching!

Stable Latent Reasoning --- Enhancing Inference in Large Language Models through Iterative Latent Space Refinement

Claim-Level Reliability Assessment for Efficient Test-Time Reasoning

Experiments on test-time scaling approaches for reasoning LM's to enforce better <think> or <wait> capabilities.