5 ms·
"LLM’s can’t scope themselves to be strictly true or accurate" This isn't true though the techniques to do so are 1. Not as yet widespread 2. Decrease the gene
by iandanforth 4y ago
"LLM’s can’t scope themselves to be strictly true or accurate"
This isn't true though the techniques to do so are 1. Not as yet widespread 2. Decrease the generality of the model and its perceived effectiveness.
- shawntan 4y agoI'm interested to hear what these techniques are. Decreasing the generality will help, but I fail to see how that scopes the output. At best that mitigates the errors to an extent.
- 256lie 4y agoconformal prediction
- shawntan 4y agoPredicting a set of answers to some confidence interval would still result in hallucinated answers.
- HDThoreaun 4y agolow probability answers get shot to a human and reviewed for model improvement.
- fulafel 4y agoRequiring the answers to automatically verifiable, or having answers be inputs to a reliable query system?
- le-mark 4y agoI’d rather use the reliable query system!