Latest

Every published item, in order · 9845 items
Oct 30
Tomorrow 08:00, the daily lands in your inboxTelegram channel or email. Same screening as this page, no ads.
PapersBuildersSelected09:50
TIDE Speeds FSDP2 Training with Zero-Copy Collectives on 64 GPUs

The hidden cost in FSDP training is often layout conversion and collective interference rather than compute, and TIDE shows both can be largely eliminated with standard NCCL APIs rather than custom hardware. Teams running large-model training on multi-node clusters can adopt it to reclaim step time and GPU memory without changing FSDP2's per-parameter kernels. The mixed-precision bf16-to-fp32 reduce-scatter is a capability stock NCCL lacks, making the implementation details directly reusable.

arXiv cs.LG@Zejian Wang et al. (14)Source
PapersBuildersOfficialSelected00:23
Meta Deploys JiTTesting LLM Bug-Catching System to Production

JiTTesting shifts test generation from pre-planned suites to on-demand tests written exactly when a risky diff appears, with humans confirming rather than authoring. Engineering teams running large monorepos can study Meta's human-on-the-loop design as a template for adding LLM-generated checks without letting unreviewed code gate merges. The paper gives concrete deployment lessons rather than a lab-only prototype.

Primary sourceMeta Engineering BlogSource
Oct 29
PapersBuildersSelected02:38
Google DeepMind's CALLIOPE maps LLM topics under differential privacy

Keyword-based pipelines can beat embeddings when privacy noise is heavy, a counterintuitive result for anyone building usage analytics on sensitive logs. Teams deploying private analytics should budget for the 2-5x token overhead and prototype with LLM-generated topic trees instead of embedding clusters. No regional angle is given, so the practical comparison is with URANIA, the embedding-based baseline it outperformed.

rohanpaul_ai@rohanpaul_aiSource
PapersBuildersSelected00:10
ARC-AGI scores follow a power curve across benchmarks

A 4.53 exponent means small gains on ARC-AGI-1 translate into large, nonlinear gaps on ARC-AGI-2, so headline ties on the older benchmark hide real capability differences. Teams benchmarking reasoning models should prioritise ARC-AGI-2 results when comparing candidates. The curve also lets practitioners extrapolate v2 scores from widely reported v1 numbers without re-running evaluations.

Official accountARC Prize@arcprizeSource

Only the latest 2000 items are listed; for older ones see the daily archive