Hacker News new | ask | show | jobs
by remify 307 days ago
LLMs fall short on most edge cases
1 comments

which would be explained that those contribute very little to weighting, and so like extrapolating beyond the last end-point, errors accumulate significantly.