Show HN: We matched full-context recall on ~1% of the tokens (open benchmark)

2 pointsposted 7 hours ago
by compresh
(github.com)

1 Comments