Unverified50% confidenceFactExact time
The poster noted that the leaderboard image in the post is somewhat outdated, and the specific figures may reflect delays or differences in how they are measured.
1
Sources
50%
Confidence
Short-term (~14 days)
Relevance
10/4/2026
First Seen
Valid until: 10/18/2026
Sources
Related Entities
Related Claims
UnverifiedStale context may overwrite more recent decisions, causing behavioral regression.66% similarUnverifiedFew-shot prompts are themselves much longer, and the mere difference in length is sufficient to shift the model's representations, meaning that previously measured drift confounded length noise unrelated to the content of the examples.65% similarUnverifiedCommonly used aggregate metrics (such as overall accuracy and evidence coverage) miss claim-level patterns, and macro-level numbers can become disconnected from actual per-claim changes.65% similarUnverifiedThe current version still falls short on details such as wiring aesthetics and avoiding overlaps, and often requires human feedback to iterate and improve.65% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/973937API
curl https://kongchang.com/api/v1/knowledge/claims/973937MCP
get_claim(id=973937)