All Updates

Microsoft Builds Own AI to Cut OpenAI Reliance

Microsoft is developing powerful in-house AI models to achieve self-sufficiency and reduce dependence on OpenAI. This strategic shift follows a relationship reorganization in October last year. The company is now independently building its most advanced AI technology.

#research#microsoft#ai

#new-feature#claude#free-tier

Anthropic Expands Claude Free Features vs ChatGPT Ads

Anthropic is opening premium features to free Claude users, including file creation/editing, third-party connectors, and Skills. This counters OpenAI's introduction of ads in free and low-tier ChatGPT. The strategy emphasizes ad-free experience with enhanced free capabilities.

#launch#google#gemini-31-pro

Google Gemini 3.1 Pro Launch Imminent

Google appears set to release Gemini 3.1 Pro soon, with model references already spotted in related arenas. This follows recent launches like Zhipu's open-source GLM-5 and DeepSeek's upgraded model with larger context window.

#update#google-gemini#content-policy

Gemini Blocks Disney Content Post-IP Claim

Google Gemini and related tools now refuse Disney character generation requests after Disney's IP infringement notice. The update rolled out about two months after Disney's December cease-and-desist letter.

Wavelet Flows Speed Universe Reconstruction

Cosmo3DFlow uses 3D wavelet transform and flow matching for efficient cosmological inference from N-body simulations. Addresses sparsity via spectral compression, enabling 50x faster sampling than diffusion models. Samples initial conditions in seconds at 128^3 resolution.

#research#cosmo3dflow#v1

VulReaD: KG-Guided Vulnerability Reasoning

VulReaD uses a security knowledge graph and teacher LLM for CWE-consistent vulnerability detection beyond binary classification. Student models are fine-tuned with ORPO for taxonomy-aligned reasoning. Boosts F1 scores significantly on real datasets.

#research#vulread#v1

VLM-Enhanced RL for Autonomous Driving

Found-RL integrates foundation models into RL for end-to-end driving via async batch inference to cut latency. Distills VLM guidance using VMR, AWAG; CLIP rewards shaped by conditional alignment. Lightweight policy matches VLM perf at 500 FPS.

#research#found-rl#v1

Visual Jailbreaks Hit Image Editors

Vision-Centric Jailbreak Attack (VJA) uses visual inputs to bypass safety in image editing models. IESBench benchmark tests vulnerabilities with up to 80.9% success rates. A training-free defense via multimodal reasoning mitigates risks effectively.

#security#vja#iesbench

VESPO Stabilizes Off-Policy LLM Training

VESPO introduces variational sequence-level soft policy optimization to tackle training instability in RL for LLMs caused by policy staleness and async execution. It derives a closed-form reshaping kernel for importance weights without length normalization. Experiments demonstrate stable training up to 64x staleness on math benchmarks.

#research#vespo#v1

Versor Revolutionizes Geometric Sequences

Versor uses Conformal Geometric Algebra (CGA) for sequence modeling with SE(3)-equivariance. Outperforms Transformers on N-body dynamics, topology, and benchmarks with fewer parameters. Offers linear complexity and interpretability via rotors.

#research#versor#v1

V-STAR: Value-Guided RecSys Sampling

V-STAR addresses probability-reward mismatch in generative recsys via value-guided decoding and sibling-relative RL. VED efficiently explores high-potential prefixes; Sibling-GRPO focuses on decisive branches. Outperforms baselines in accuracy and diversity.

#research#v-star#v1

Universal Multimodal Immune System Model

EVA is a cross-species, multimodal foundation model harmonizing transcriptomics and histology for immunology. It shows scaling laws and SOTA on 39 tasks from discovery to clinical trials. Open version released for transcriptomics research.

#research#eva#v1