7 ms·
LLMs fall short on most edge cases
by remify 1y ago
LLMs fall short on most edge cases
- stargrazer 1y agowhich would be explained that those contribute very little to weighting, and so like extrapolating beyond the last end-point, errors accumulate significantly.