Content pfp
Content
@
0 reply
0 recast
0 reaction
Varun Srinivasan pfp
Varun Srinivasan
@v
Is there a credible solution to the LLM hallucination problem? Any interesting research papers or discussions on this?
9 replies
3 recasts
58 reactions
ȷď𝐛𝐛 pfp
ȷď𝐛𝐛
@jenna
Curious whether/how for you “hallucination” distinct from a broad “wrong answer” category (delivered confidently in both cases ofc lol)
1 reply
0 recast
0 reaction
jj 🛟 pfp
jj 🛟
@jj
https://openai.com/index/webgpt/ It’s been worked on since 4 years ago It’s come a long way
0 reply
0 recast
3 reactions
Shashank pfp
Shashank
@0xshash
amazon recommends using RAG on knowledge sources to get a hallucination score https://aws.amazon.com/blogs/machine-learning/reducing-hallucinations-in-large-language-models-with-custom-intervention-using-amazon-bedrock-agents/ research paper: https://www.amazon.science/publications/hallucination-detection-in-llm-enriched-product-listings
0 reply
0 recast
0 reaction
mk pfp
mk
@mk
Seems like since it happens rarely, multiple LLMs working together should be able to detect them. I don’t understand why we keep applying one LLM when our own minds use multiple models.
1 reply
0 recast
0 reaction
Vinay Vasanji pfp
Vinay Vasanji
@vinayvasanji.eth
Perhaps something directionally to do with reasoning capability https://www.machinesforhumans.com/post/unreasonable-ai---the-difference-between-large-language-models-llms-and-human-reasoning
0 reply
0 recast
0 reaction
Stephan pfp
Stephan
@stephancill
How prevalent is it in coding applications? Feels like I never experience LLM hallucination but coding is pretty much my only use case
1 reply
0 recast
0 reaction
Dean Pierce 👨‍💻🌎🌍 pfp
Dean Pierce 👨‍💻🌎🌍
@deanpierce.eth
The main solution is for the agent to provide reputable sources for any claims, but it's usually cheaper and easier for the human operator to do the fact checking. Letting an agent randomly browse around online to do fact checking is slow and expensive, so it's not the default, but it's totally possible.
0 reply
0 recast
0 reaction
Idan Levin 🎩 pfp
Idan Levin 🎩
@idanlevin
there is a cocept of truth orcale (sort of fact checking service) that some are exploring
0 reply
0 recast
0 reaction
Mkkstacks pfp
Mkkstacks
@mkkstacks
@sophia-indrajaal
1 reply
0 recast
1 reaction