Fictional EN/TA facts, capitals pilot, and every run log, generation and blind audit (Lexsi hackathon, Track 2b).
Vinod Anbalagan
vinod-anbalagan
AI & ML interests
ML engineer and independent researcher working on representation learning, multimodal AI, world models, and efficient learning. I build models, datasets, evaluation systems, and research artifacts in JAX, PyTorch, and Python, with a focus on understanding what models learn and how learning objectives, architectures, and constraints shape their representations.
Recent Activity
published a Space 5 days ago
vinod-anbalagan/where-forgotten-knowledge-goes published a dataset 5 days ago
vinod-anbalagan/unlearn-audit-tiny-aya-logs published a dataset 5 days ago
vinod-anbalagan/tiny-aya-fictional-facts-en-ta