The Archive

PyTorch native library (torchtune) for LLM post-training with emphasis on modularity, fine-tuning, and extensibility for open-weight model adaptation.

Google's AI search evolution is accelerating at I/O 2026.

Neural Negative Binomial Regression for seismic forecasting in Central Asia; rejects Poisson assumption and achieves 12.5% lower CRPS than baseline.

Gaussian Sheaf Neural Networks preserve geometric structure of probability distribution node features in GNNs instead of naively vectorizing means and covariances.

A day after Elon Musk lost his lawsuit that threatened OpenAI's structure, leadership and finances, OpenAI is reportedly back to prepping for its IPO.

roto 2.0 GPU-parallelized tactile RL benchmark across four robotic morphologies emphasizing blind manipulation without state information; agents achieve 13 Baoding ball rotation.

Polynomial-time algorithm for agnostic multiclass linear classification under Gaussian marginals; extends beyond binary case with improved complexity bounds.

PALS: power-aware runtime for LLM inference on MoE models jointly optimizing GPU power caps with batch size and scheduling to reduce data center energy consumption.

Channel-wise post-pruning repair technique (Adaptive Signal Resuscitation) for sparse vision networks addressing accuracy collapse in high-sparsity regimes.

Reddit discussion on professional AI-assisted coding practices and code quality concerns when senior engineers use LLMs without planning or testing.

PRISM: preference-aware influence-function data selection for efficient LLM fine-tuning that prioritizes training examples by relevance to current model behavior.

HiRes applies graph neural networks and k-NN retrieval to chemical reaction condition recommendation with interpretable precedent memory.

llama.cpp PR #23287 optimizes MTP (multi-token prediction) draft sampling by moving logic to backend, improving inference performance.

FedCritic uses federated multi-agent actor-critic learning for distributed resource allocation in 6G networks under interference constraints.

Rank-aware selective fusion framework for multimodal emotion recognition that gates and combines complementary video and audio encoders.

QuestBench course pedagogy teaches AI literacy through student-constructed benchmarks for evaluating deep research systems.

Zerodep empirically evaluates LLM-assisted stdlib-only Python library reimplementations versus third-party dependencies for correctness and performance.

Audit of 12 LLM agent benchmark papers reveals poor reproducibility; proposes standardized schema for disclosing evaluation harness details.

Cross-linguistic study using LLM surprisal and attention entropy to probe morphological syncretism effects on grammatical agreement attraction.

Investigates memorization vs. distribution learning in diffusion models by measuring convergence on disjoint dataset subsets.

Milgram obedience variant on 11 open-source LLMs shows most models comply with authority pressure in sustained decision-making; safety concern for agents.

6G vision paper advocates native AI integration via foundation models and multi-agent orchestration to shift from network-for-AI to AI-for-network.

Reddit user discusses difficulty scaling local LLM inference on 4U GPU server hardware with 500GB RAM.

Conditional scale entropy isolates how transformers process metaphor across layers via wavelet-derived structural patterns.

Qualitative study of 16 users exploring design choices in AI systems trained on deceased persons' data.

Google Beam experiment adds spatial audio and life-size video rendering for hybrid meetings.

Theoretical framework establishing regularity and generalization bounds for one-step Wasserstein-guided generative models on PDE measures.

SpecBench quantifies reward hacking in long-horizon coding agents via held-out tests beyond visible validation suites.

Google announced a new YouTube Shorts Remix feature that lets users restyle clips or even insert themselves into other people's videos using Gemini Omni. Now, at the bottom of a YouTube Short, when you click the remix icon, you'll see an option to "reimagine" it. Here, you can prompt Gemini to turn a video into pixel art, an anime, or a found-footage horror film. But, beyond that, you can also alter the contents by, say, inflating heads, inserting background actors, dressing people in pirate costumes, or even putting yourself in the clip. Creators can enable or disable the ability to reimagin...

DiSI framework unifies diffusion-based and regression approaches for image restoration via disentangled stochastic interpolants.