bytedance / bytedance/Dolphin

幻觉问题异常严重

Open
#153 1 comment 0 reactions 0 assignees View on GitHub
Dominant language
Python
Stars
9.1k
Forks
777
PR merge metrics
No merged PRs in 30d

Description

1. 之前刚发布时就关注了,虽然速度很快,但是幻觉特别严重,总是无中生有,完全无法使用
2. 前几天发现发布了Dolphin-1.5,今天试了下,相比之前幻觉稍微减少了那么一点点,但还有无法使用,而且之前的还能识别复选框(虽然效果不好),1.5版本就直接忽略了复选框。

希望能重点解决一下幻觉问题,不然是完全无法使用的,其它OCR模型现在基本都没遇到幻觉。

Contributor guide

No contributing guide indexed for this repository

Research direction

Start by reproducing the reported behavior with Dolphin-1.5 on documents that contain checkboxes, and compare its output with the earlier version. Investigate the hallucinated content and ignored checkboxes; done would require substantially reducing hallucinations and preserving checkbox recognition, but the issue names no files or tests.

Written by the indexing model from the issue text.

Assessment

Tech stack
python
Domain
computer-vision, machine-learning
Issue type
Bug
Difficulty
5/5
Estimated time
Over a week
Activity status
Stale
Clarity
Needs clarification
Newbie friendliness
25/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.