Learning Functional Subspaces for Neural Network Compression
Learnable Subspace Projections compress LLMs end-to-end, beating baselines at 70% compression on Llama-2-7B.
Stories per day · 30 days
2 RSS
3 stories, newest first · one card per story
Stories tagged #llama-2
Safer Content or Firmer Refusals? A Hybrid Perturbation Defense for Alignment under Harmful Fine-tuning
VaccineBooster lowers flagged harmful output after poisoned fine-tuning but can weaken explicit refusals.
Seal, Then Sample: Sampled Layerwise Proofs for Verifiable LLM Inference from GPT-2 to 70B
Researchers propose Sampled Layerwise Proofs to audit outsourced LLM inference through Llama-2-70B.