This model keep telling lies
#61
by
awesome747
- opened
On the surface, it looks promising and faster. It shows all the good plan.
When I examine the execution, none of them happened. I need to go to the directory and check the file it claims to have created and all are non-existences and making things up.
When I ask what model you are, zai even said I am Claude, an AI assistant made by Anthropic!
Did you plug it into an agentic system with the actual tooling to do file operations?
Also, practically all local models do that (the wrong name thing), LLM generally have no "awareness" of what they are. All this show is that Claude's output was part of the training data (which is common practice).