News for hackers. Stuff that matters.

A comment on Replace OCR with Vision Language Models. Parent comment

Re:Replace OCR with Vision Language Models by _delirium on Thursday February 27, 2025 @12:45PM (#43196612)
Yes, there’s research showing that models’ self-assessment of probabilities (when you ask them via prompting) don’t even match the same models’ actual probabilities, in cases where you can measure the probabilities directly (e.g. by looking at the logits): https://arxiv.org/abs/2305.13264
[ Parent | on HN ]
Re:Replace OCR with Vision Language Models by anon291 on Friday February 28, 2025 @12:06AM (#43201901)
Logits are not probabilities... at least not in the way you understand probability. Probabilities mathematically are anything that broadly behaves like a probability, whereas colloquially probabilities represent the likelihood or the preponderance of a particular phenomenon. Logits are not either of those.
[ Parent | on HN ]
Re:Replace OCR with Vision Language Models by _delirium on Friday February 28, 2025 @03:20AM (#43203115)
The probability of token generation is a function of the logits. Do you have an actual point related to the linked paper?
[ Parent | on HN ]
Re:Replace OCR with Vision Language Models by anon291 on Friday February 28, 2025 @02:32PM (#43209372)
That is one way of sampling tokens. It is not the only way. Logits do not map neatly to belief, although it is convenient to behave as if they do
[ Parent | on HN ]