Fetching from the wire…
01
02
03
04
05
06
07
08
Research2026-08-12 · source-backed
Sampled softmax cuts the O(nK) memory of full-vocabulary classification to O(nk), but for fixed budget B = n·k it's been unclear whether to buy batch or negatives (arXiv 2608.11061). Analyzing convergence under standard smoothness and variance assumptions, the fastest convergence comes from n ~ B, k ~ 1. Confirmed on synthetic experiments and four real sequential recommendation benchmarks including MovieLens-20M. Rare thing: a paper that ends in a rule you can apply without reading it twice.
Each link below shares sources, entities, or timing with this story.
The method stores hardware kernel optimization trajectories, with correctness and performance feedback, in an Experience Graph Memory that preserves decision order, observed outcomes, and abandoned branches, then retrieves under a fixed token budget (arXiv 2608.25570). Under t...
Standard evaluation of frozen-embedding style classification uses random splits where works by the same artist appear on both sides (arXiv 2608.14435). Under an artist-disjoint protocol on 320 paintings across four twentieth-century movements, 5-NN accuracy falls ten points, a...
Nobody had quantified this mismatch: KV eviction algorithms like H2O and Scissorhands work at token granularity while PagedAttention manages fixed-size blocks, so evicted tokens leave intra-block holes that never get reclaimed. vToken's token-table indirection layer with async...
arXiv 2607.29390 shows preprocessing and event-labeling methodology varies substantially paper to paper on two popular datasets, quietly deciding how fair the resulting evaluation is. Under grounded policies, three well-known detectors score differently than published. If you...
Fourati, Schütze, Hüllermeier and Gurevych challenge the assumption that humans stay in the loop only because AI isn't capable enough yet, identifying three durable grounds: complementarity, normative/developmental value, and the one they weight most heavily, target emergence...
Most benchmarks test single-episode solving, and memory benchmarks test fact retention. Neither checks procedural reuse, whether an agent can convert a solved session into a reusable search/debug/verify routine (arXiv). Under a Train/Extract/Test protocol with held-out tasks,...
MindPattern daily
One email a day at 7 AM. Sources and a take on every story. Unsubscribe anytime.