Catastrophic Failure of LLM Unlearning via Quantization — mentioned by sigbottle

Thu, 09 Oct 2025 18:02:05 GMT

Not necessarily? The way these models are trained suggests "more good data is more good". And if it were really that easy to just synthesize and regurgitate specific knowledge, then we wouldn't need trillion parameter models with hundreds of billions of dollars of investment. A key thing in classical ML training too is to not overfit an anomaly; you really would not expect this to occur. Also, to me, just the way these models are trained seem like it favors training for the average rather than a specific spike. A middle ground might be, "Learning to spit arbitrary text at a poisoned token is…

Paper and discussion history. A mention is not an endorsement.

Catastrophic Failure of LLM Unlearning via Quantization — HN submission

Mon, 04 Nov 2024 02:52:14 GMT

An embarrassingly simple approach to recover unlearned knowledge for LLMs

259 points; 121 discussion replies on HN. Paper and source history. A submission is not an endorsement.

Catastrophic Failure of LLM Unlearning via Quantization · HackerPapers discussions

Catastrophic Failure of LLM Unlearning via Quantization — mentioned by sigbottle

Catastrophic Failure of LLM Unlearning via Quantization — HN submission