3 ms·
why do these LLMs struggle so hard to be concise? is this because the Alpaca dataset is pretty wordy? I'd be interested in trying to run fine-tuning to make it
by sterlind 4y ago
why do these LLMs struggle so hard to be concise? is this because the Alpaca dataset is pretty wordy? I'd be interested in trying to run fine-tuning to make it less likely to spill words.
- madmads 4y agoI heard somewhere that during the RHLF process of ChatGPT, people tended to rate longer responses higher. It might be something similar going on here.
- ghotli 4y agoIn my experience so far GPT-4 will most certainly be concise if you ask it to be. Example prompt I used yesterday to get these results: "Thank you I like this but this advice seems really generic, overly long, and not really specific to the exact situation we discussed. Can we simplify this down to ..." Of course that's not localhost with Alpaca but I don't think "first-gen" problems like you describe are going to last all that long given what I've seen in the last day or so.
- visarga 4y agoConciseness is regulated from the 3rd stage of training, RLHF. Alpaca only had the first two stages.