#logprobs

Ingosstrakh Develops Confidence Scoring for LLM Document Processing

Ingosstrakh's AI team is developing a method to detect errors in multimodal LLM document processing by analyzing token logprobs to calculate confidence scores.