Editor's Picks

Latest

View all

CALM Cuts Trial-Estimate Variance While Preserving Validity

LLM counterfactual predictions are calibrated and cross-fitted into causal estimators, lowering variance across 300 BRIGHTEN-based simulation runs.

Aug 29, 20265 min2510.05545

Three Exceptional Values Do Not Force Bounded Type

A symmetric meromorphic construction places the preimages of 0, 1, and infinity on the real axis while making every other point divisor non-Blaschke in both half-planes.

Aug 29, 20264 min2608.26062

Adhesion Changes Epithelial Fluidity Without Reshaping Cells

By adding adhesion-dependent intercellular friction to a vertex model, the analysis explains fluidity changes across five DECMA-1 concentrations without a shape-index shift.

Aug 28, 20265 min2603.05548

TurnBench Exposes Turn-Taking Failures Across Conversation Types

A 30-hour triple-annotated benchmark separates end-of-turn and interruption decisions, showing that low-latency interruption detection still produces excessive false positives.

Aug 27, 20264 min2608.25218

Audit Cuts Green Patent Count by a Quarter

An error-as-signal audit combines PatentSBERTa disagreements with two-model climate-function judgments, reducing the Y02 estimate from 592,387 to 441,468 patents.

Aug 27, 20265 min2608.23420

Calabi Curvature Operator Characterizes Complex Projective Space

A new Bochner identity turns a Calabi-operator positivity condition into a rigidity result for Kähler–Einstein manifolds.

Aug 26, 20263 min2608.23778

FLARE Extends Laboratory Access to Multiple X-Line Reconnection

A larger, modular reconnection experiment combines segmented coils, independent ohmic heating, and upgraded diagnostics to target $S\sim10^5$ and $\lambda\sim10^3$.

Aug 26, 20265 min2608.17332

End-to-End Audio Generation Cuts Speech Errors

An LLM predicts audio latents autoregressively while per-token flow matching generates variable-length scenes, reducing English Seed-TTS WER from 12.15% to 2.79%.

Aug 25, 20264 min2608.11804 Code available

Mobius Separates Knowledge From Reasoning for Faster Inference

A shared feed-forward memory and iterative attention reasoners retain comparable downstream scores while reducing continual-pretrained 35B-model inference time by nearly 4×.

Aug 25, 20264 min2608.14290

Better Netflix Recommendations Shift Viewing Toward the Middle Tail

An 8.5-million-user Netflix experiment finds that improved ranking raises engagement while moving recommendation-driven consumption from superstars toward moderately popular titles.

Aug 24, 20264 min2608.21274

Get the papers that matter, weekly

One email a week. The most significant new research, distilled. No spam. Unsubscribe anytime.