Extracting memorized pieces of (copyrighted) books from open-weight language models · HackerPapers discussions

Extracting memorized pieces of (copyrighted) books from open-weight language models — mentioned by grey-area

Thu, 26 Feb 2026 09:44:56 GMT

I may be using the wrong terms, my impression was: 1. Weights in the model are created by ingesting the corpus 2. Techniques like reinforcement learning, alignment etc are used to adjust those weights before model release 3. The model is used and more context injected which then affects which words it will choose, though it is still heavily biased by the corpus and training. That could be way off base though, I'd welcome correction on that. The point I was trying to make though was that they do more than predict next word based on just one set of data. Their weights can encode entire passages…

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

Extracting memorized pieces of (copyrighted) books from open-weight language models — HN submission

Mon, 16 Jun 2025 17:41:27 GMT

Extracting memorized pieces of books from open-weight language models

109 points; 109 discussion replies on HN. Paper and source history. A submission is not an endorsement.

Extracting memorized pieces of (copyrighted) books from open-weight language models — mentioned by aspenmayer

Sun, 15 Jun 2025 17:18:51 GMT

As there is no reason to believe that Harry Potter is axiomatic to our culture in the way that other concepts are, it is strange to me that the LLMs are able to respond in this way, and not at all expected. Why do you think this outcome is expected? Are the LLMs somehow encoding the same content in such a way that they can be prompted to decode it? Does it matter legally how LLMs are doing what they do technically? This is pertinent to the court case that Meta is currently party to. https://en.wikipedia.org/wiki/Artificial_intelligence_and_co... > See for example OpenAI's comment in the year…

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

Extracting memorized pieces of (copyrighted) books from open-weight language models — HN submission

Tue, 27 May 2025 17:23:38 GMT

Extracting memorized pieces of books from open-weight language models

2 points; 0 discussion replies on HN. Paper and source history. A submission is not an endorsement.