Skip to main content
← Lab

attnoracle

null resultPythonInference & servingJuly 11, 2026

the question

Is attention-mass the right objective for KV-cache eviction?

what came out

No, and neither is the popular value-norm fix. An exact brute-force-optimal K-subset oracle on Qwen2.5 shows value-norm (w*||v||) is a trap (+55% downstream KL), while mass isn't globally optimal either -- the gap to optimum grows with context. All five pre-registered predictions were falsified.

method & receipts

  • Result: null result — an honest negative, reported as it came out.
  • Pre-registered the prediction was written down and committed before the run.

→ read the code and re-run it

github.com/v-code01/attnoracle