Beyond the Black Box: A Statistical Model for LLM Reasoning and Inference · HackerPapers discussions

Beyond the Black Box: A Statistical Model for LLM Reasoning and Inference — mentioned by zipfcharge

Wed, 08 May 2024 23:51:46 GMT

It's because an LLM is essentially a probability matrix. You type a prompt, then it calculates what's the probability of getting a next word and so on, eventually forming a sentence. The probability learned is based on the training data. Because of the underlying probability model, it's not going to be 100% deterministic. Plus a model like ChatGPT purposefully have "temperature" parameter that will further add randomisation to the whole process. My answer is based on this paper if you're interested to read more: The Matrix: A Bayesian learning model for LLMs, https://arxiv.org/abs/2402.03175

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

Beyond the Black Box: A Statistical Model for LLM Reasoning and Inference — HN submission

Sat, 04 May 2024 09:27:45 GMT

The Matrix: A Bayesian learning model for LLMs

139 points; 10 discussion replies on HN. Paper and source history. A submission is not an endorsement.

Beyond the Black Box: A Statistical Model for LLM Reasoning and Inference — HN submission

Fri, 03 May 2024 17:10:29 GMT

The Matrix: A Bayesian learning model for LLMs

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

Beyond the Black Box: A Statistical Model for LLM Reasoning and Inference — HN submission

Sun, 28 Apr 2024 15:35:55 GMT

The Matrix: A Bayesian Learning Model for LLMs

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