FineVision: Open Data is All You Need
š
215
A new open-source dataset for training VLMs
Jan Erik van Woerden
ijanerik
Ā·
AI & ML interests
None yet
Recent Activity
liked
a Space
3 months ago
HuggingFaceM4/FineVision
Organizations
None yet
liked
a
Space
3 months ago
upvoted
a
paper
6 months ago
reacted to
m-ric's
post with š„
8 months ago
Post
2740
A new research paper from KAIST builds on smolagents to push boundaries of distillation š„³
ā”ļø "Distilling LLM Agent into Small Models with Retrieval and Code Tools" teaches that, when trying to distil reasoning capability from a strong LLM ("teacher") into a smaller one ("student"), it's much better to use Agent traces than CoT traces.
Advantages are:
1. Improved generalization
Intuitively, this is because your agent can encounter more "surprising" results by interacting with its environment : for example, a web research called by the LLM teacher in agent mode can bring results that the LLM teacher would not have generated in CoT.
2. Reduce hallucinations
The trace won't hallucinate tool call outputs!
Thank you @akseljoonas for mentioning this paper!
ā”ļø "Distilling LLM Agent into Small Models with Retrieval and Code Tools" teaches that, when trying to distil reasoning capability from a strong LLM ("teacher") into a smaller one ("student"), it's much better to use Agent traces than CoT traces.
Advantages are:
1. Improved generalization
Intuitively, this is because your agent can encounter more "surprising" results by interacting with its environment : for example, a web research called by the LLM teacher in agent mode can bring results that the LLM teacher would not have generated in CoT.
2. Reduce hallucinations
The trace won't hallucinate tool call outputs!
Thank you @akseljoonas for mentioning this paper!
liked
a
model
over 2 years ago
