The Archive

VocalGrad benchmark for isolating acoustic perception failures in large audio language models, distinguishing paralinguistic understanding from reasoning.

Tool Recommendation system deployed for multi-step agents surfaces 8,200 tools through fixed interface, solving context-window bottleneck at scale.

Multi-task active learning framework using orthogonal decomposition instead of linear scalarization to resolve gradient conflicts across objectives.

Ablation study isolates architecture vs. loss contributions in competing-risks survival analysis, decoupling confounded deep learning design choices.

NestedVLA consolidates multi-task robotic skills via hypernetwork-based parameter generation instead of parameter merging.

ADT decouples token granularity from reconstruction resolution in 3D molecular diffusion tokenization for proteins and small molecules.

Counterfactual pairing cycles benchmark method isolates relational OOD errors in vision-language models by controlling caption style and image statistics.

LangNCO: framework translating natural-language constraints to inference-time interventions for neural combinatorial solvers.

AMEND: inference-time multi-agent framework for text detoxification with constraint-enforced semantic preservation.

Grupo Boticário deploys proprietary agent platform resolving 80% of reseller requests autonomously, shifting sales team to advisory roles.

Time series language models benefit from complementary dual representations: line plots for absolute values, coordinate frames for variations.

EliteReplay combines GRPO with quality-diversity search to preserve solution diversity in RL post-training rather than converging to single policies.

Topology-aware loss functions for wind-field super-resolution favor different model architectures than pointwise metrics.

Training-free layer-wise consolidation fuses specialist models into one student using only forward passes, eliminating distillation's training cost.

Claude Opus 4.8 and Codex GPT Sol 5.6 fail at reverse-engineering iOS protobuf schemas with high error rates; LLMs unreliable for forensic analysis.

BiShield-TEE identifies vulnerability in unilateral weight obfuscation for on-device LLMs via Solution Determinacy Property, exposing model structure.

Proposes Function-isolated Memory-Efficient Optimization (CiME) localizing low-overlap functional subnetworks in LLMs for continual learning without catastrophic forgetting.

ISW-Agent integrates LLM with n8n workflows for internal solitary wave remote sensing; achieves 94.2% task-match rate on 120 Chinese instructions.

The founders detailed how their startup is developing AI that can design chips, learn from each iteration, and improve subsequent designs, all with the goal of automating and accelerating the chip design process.

Enterprise strategies for managing agent proliferation and leveraging controlled duplication to discover effective agentic solutions.

New York Attorney General Letitia James revealed a years-long multi-state investigation against Character.AI as they now seek to require protections on consumer data sold through its bankruptcy proceeding. James is among 40 bipartisan AGs who filed a notice of interest in Character.AI's bankruptcy case "to put all parties on notice of the regulatory issues that will need to be addressed in any sale or transfer of the Debtor's business or assets." Character.AI announced Tuesday it was filing for Chapter 11 and selling some assets to Disney. But the AGs warn there could be serious risk involved...

Mixture-of-experts architecture using Hodge-Laplacian gating partitions PDE solutions by spectral scale, improving surrogate model performance on unstructured meshes.

PerSo benchmark evaluates LLM social intelligence in real-world persuasion via both cognitive and behavioral aspects beyond synthetic scenario isolation.

VascuSR combines iterative diffusion with 3D vascular guidance for anatomically-plausible CT super-resolution avoiding spurious vessel hallucination.

Study shows LLM agents and scientific verifiers tacitly select single interpretations of underspecified claims, causing benchmark disagreements that hide interpretation variance.

Old-school physics-based weather forecasting and even new AI models typically become less reliable beyond two weeks, a problem Microsoft claims to have solved using machine learning. Researchers with the company say they've found a new, cost-effective way to make more accurate weather forecasts two to six weeks in advance - allowing farmers and emergency responders to plan ahead with greater certainty. The findings were published today in the journal Nature, and the tool is now open-source software. Microsoft joins Google, Nvidia, and other companies developing AI tools for weather forecastin...

Cortex consolidates context-specified knowledge into LLM parameters via prospective queries and distillation, unifying context editing, skill learning, and behavioral steering.

LAGO meta-agent framework treats specialized agents as reusable options; resolves 58.6% of SWE-bench tasks with frozen Qwen3.5-35B.

CTIA-ODE applies continuous-time attention within ODE solvers for multivariate time series forecasting with exogenous variables.