Harmonizing Multi-Objective LLM Unlearning via Unified Domain Representation and Bidirectional Logit Distillation

Create a modular unlearning library that implements bidirectional logit distillation. It should help researchers strip specific knowledge from models while retaining general reasoning ability.

Suggested repo: forget-logits

"Surgical LLM unlearning with minimal utility loss."

Estimated effort: 30h