The slow normalization of benchmark contamination
The spec is settled. The implementation is another matter.
The conversation around benchmark contamination keeps returning to the same trade-off, usually because nobody wants to price it.
Subscriber story
Continue reading with a subscription
This piece is part of the subscriber archive. Subscribe to the daily brief for full access — one email, no card, cancel any time.
Already a subscriber?
#evaluation#infrastructure#inference
More in this section
ai
Why small models are winning on-device
Four-bit quantization plus better distillation is making 3B models genuinely useful in production.
🔒aiSubscriber
Synthetic data is hitting a wall nobody wants to name
Model collapse is real, but the more immediate problem is boring: attribution and licensing.
🔒aiSubscriber
The open-weights licence split is becoming permanent
Two families of licence, two incompatible expectations, and a lot of downstream confusion.