work VLM Failure Mode Analysis Do vision-language models trust the image, or follow misleading language priors? Targeted Latent Adversarial Training — Reproduction Laptop-scale reproduction of latent-space adversarial robustness for LLMs. LogLLM — Reproduction Reproducing LLM-based log anomaly detection, with reproducibility and robustness fixes. MERINA — Reproduction & Adaptation Analysis Meta-RL adaptive bitrate streaming, with cross-dataset and Puffer adaptation analysis. Minimal MERINA+ A compact, readable reimplementation of meta-RL adaptive bitrate streaming. Imitation Learning for Continuous Control Behavioral Cloning vs. DAgger on MuJoCo locomotion, and when cloning breaks. fun Mirror-Symmetry Detection Net A from-scratch NumPy reproduction of the classic 1986 backprop symmetry network. CS224R — Deep RL Homework Coursework on deep reinforcement learning (private repository).