Subscribe
Category / page 14

Models & Assistants

1121 articles

Articles, page 14

261-280 of 1121

  1. Cohere Releases Parse 5: 2.3B Vision-Language Model for Enterprise Document Processing at $1.50 Per 1,000 Pages

    Cohere has released Parse 5, a specialized 2.3-billion-parameter vision-language model (VLM) engineered to convert complex enterprise documents into structured Markdown and HTML blocks in a single inference pass. Priced at $1.50 per 1,000 pages via API, Parse 5 targets large-scale enterprise retrieval-augmented generation (RAG) and document processing pipelines where routing millions of pages through frontier foundation models is cost-prohibitive. Single-Pass Architecture and Capabilities Tr

    AI
    2 min read

  2. Reranking Models and Late-Interaction Frameworks in Production RAG: Comparing Cohere Rerank 3.5, BGE-Reranker-v2, FlashRank, and ColBERTv2 Architecture, Cross-Encoder Latency, MaxSim Compression, and Retrieval Economics

    In production Retrieval-Augmented Generation (RAG), relying solely on first-stage bi-encoder dense embeddings or lexical BM25 search creates a structural information bottleneck. First-stage retrieval compresses an entire document passage into a single fixed-dimensional vector (such as 768, 1024, or 1536 dimensions) or an inverted term index. This independent encoding decouples the query from the passage during indexing, eliminating token-level cross-attention and causing subtle semantic nuances,

    AI
    11 min read

  3. Meta Deploys ABB and Kinova Robots to Automate Physical Hardware Tasks in AI Data Centers

    Meta has begun testing specialized robotic systems from industrial automation manufacturers to perform physical maintenance tasks across its AI data center fleet, according to internal reporting confirmed by current and former staff. The automation program, deployed across facilities including the company's Prometheus data center campus in New Albany, Ohio, uses robots to replace networking cables, power-cycle servers, reseat components, and perform routine hardware maintenance without human in

    AI
    2 min read

  4. Meta Vice President Sandhya Devanathan Joins OpenAI to Lead Southeast Asia and Australia Expansion

    OpenAI has hired Sandhya Devanathan, Meta's Vice President for India and Southeast Asia, to lead its commercial and operational expansion across Southeast Asia and Australia. Devanathan, who spent nearly a decade in senior leadership roles at Meta, will be based out of Singapore and report to Kiran Mani, OpenAI's Managing Director for the Asia-Pacific region. In her new role, she will oversee consumer product adoption, enterprise sales and partnerships, government and regulatory relations, and

    AI
    2 min read

  5. Google DeepMind Pilots Cryptographic Double-Blind AI Evaluations to Prevent Benchmark Contamination

    Google DeepMind, in collaboration with the Singapore AI Safety Institute, OpenMined, AVERI, and MLCommons, has piloted a cryptographic framework for double-blind evaluations of proprietary frontier language models. The pilot, conducted on Gemini 2.5 Flash Lite, uses hardware-isolated confidential computing to ensure that model developers cannot see evaluation prompts while evaluators cannot inspect proprietary weights or inference code. The project addresses benchmark contamination and intellec

    AI
    3 min read

  6. Mixture-of-Depths (MoD): Mathematical Foundations, Dynamic Token-Level Compute Routing, Top-k Capacity Budgeting, and FLOP-Optimal Transformer Architectures

    Standard autoregressive Transformers allocate an identical compute budget to every token in a sequence. Regardless of whether a token represents a trivial punctuation mark, a common grammatical function word, or a complex semantic reasoning step, the model applies the exact same sequence of multi-head self-attention and multilayer perceptron (MLP) operations across all $L$ layers. In Mixture-of-Depths: Dynamically allocating compute in transformer-based language models, researchers at Google De

    AI
    6 min read

  7. South Korea Launches AI for All Initiative Treating Frontier Models as Public Utilities

    South Korea's Ministry of Science and ICT (MSIT) has launched the "AI for All" initiative, a government-sponsored project aimed at deploying nationwide, free-tier access to artificial intelligence chatbots and public administrative agents. The initiative structures generative AI capabilities and agent workflows as public utilities, establishing a subsidized access model for South Korean citizens. Consortia and Public Bidding Six major commercial consortia submitted bids to operate the public

    AI
    2 min read

  8. Simple Preference Optimization (SimPO): Mathematical Foundations, Reference-Free Implicit Reward, Length Normalization, and Target Margin Dynamics

    Simple Preference Optimization (SimPO): Mathematical Foundations, Reference-Free Implicit Reward, Length Normalization, and Target Margin Dynamics Post-training preference alignment has become the definitive step in transforming raw pretrained large language models into instruction-following assistants. While Reinforcement Learning from Human Feedback (RLHF) via Proximal Policy Optimization (PPO) established the initial standard, its requirement to maintain actor, critic, reference, and reward

    AI
    11 min read

  9. OpenAI Partners with Thai Government to Launch Accelerator for Healthcare and Education Startups

    OpenAI has partnered with Thailand's Ministry of Higher Education, Science, Research and Innovation (MHESI) to launch an eight-week accelerator program supporting ten Thai startups building AI products in healthcare, wellness, and education. The initiative, announced in Bangkok, represents OpenAI's first public-private partnership focused on startup acceleration with the Thai government. Participating organizations include Thailand's National Innovation Agency (NIA), Mahidol University, and Tec

    OpenAI
    2 min read

  10. Group Relative Policy Optimization (GRPO): Mathematical Foundations, Critic-Free Advantage Estimation, Group Reward Normalization, and Reinforcement Learning with Verifiable Rewards

    Reinforcement learning from human and verifiable feedback has become the central paradigm for unlocking complex reasoning, mathematical problem solving, and autonomous code synthesis in frontier large language models. While early post-training pipelines relied heavily on Proximal Policy Optimization (PPO) or offline preference algorithms like Direct Preference Optimization (DPO), both methodologies present structural bottlenecks when scaling to long-horizon reasoning. PPO requires maintaining a

    AI
    11 min read

  11. OpenAI Tests Persistent Mode in Codex for Long-Running Autonomous AI Agents

    OpenAI is testing an execution profile termed "Persistent Mode" within its Codex agent codebase, designed to enable continuous, self-directed task execution without standard step-count timeouts or per-turn pauses. Code commits surfaced in the public repository of the Codex command-line interface indicate that the agent can proactively generate follow-up tasks, maintain state across development sessions, and continue working autonomously until explicitly halted by the user. Architecture and Re

    OpenAI
    1 min read

  12. Federal Judge Overturns Pentagon Supply Chain Risk Blacklisting of Anthropic

    A federal judge has ruled that the Trump administration broke the law when it designated Anthropic a supply chain risk, finding that the government unlawfully targeted the company for refusing to comply with Defense Department demands, The Guardian reported. U.S. District Judge Rita Lin issued a 59-page decision overturning Defense Secretary Pete Hegseth's designation and barring the federal agencies named in the lawsuit from enforcing President Trump's order to stop using Anthropic's tools. Th

    Anthropic
    2 min read

  14. Anthropic Opens 10,000 Claude Seats and Expands Grants for Academic Scientists

    Anthropic has introduced a dedicated scientific research initiative providing 10,000 subsidized Claude Team subscription seats and expanding API grant funding for academic and non-profit institutions. The program broadens the company's research support beyond initial biology use cases into mathematics, physics, engineering, and computer science. Under the new plan, accredited academic institutions and non-profit research organizations can register research groups under principal investigator ad

    AI
    2 min read

  15. Tencent Open-Sources Hy4-Preview: 770B MoE Architecture with 1M Context

    Tencent has released the open-weights preview of its next-generation foundation model, Hy4-preview, under the Apache 2.0 license. The release scales the organization's Mixture-of-Experts (MoE) line to 770 billion total parameters, activating 49 billion parameters per token across a native 1-million-token context window. Weights have been published across Hugging Face, ModelScope, GitCode, and GitHub, alongside managed API availability on Tencent Cloud TokenHub and OpenRouter. Architectural Sp

    AI
    2 min read

  16. Open-Source Embedding Models and Serving Frameworks in Production: Comparing BGE-M3, NV-Embed-v2, GTE-Qwen2, and ModernBERT-Embed Architecture, Matryoshka Projections, Context Scaling, and Serving Economics

    Open-Source Embedding Models and Serving Frameworks in Production: Comparing BGE-M3, NV-Embed-v2, GTE-Qwen2, and ModernBERT-Embed Architecture, Matryoshka Projections, Context Scaling, and Serving Economics (TEI vs. Triton vs. vLLM) Text embeddings form the indexing and retrieval foundation for production Retrieval-Augmented Generation (RAG), semantic search, and agentic memory systems. While early production architectures relied almost exclusively on closed commercial APIs (such as OpenAI's te

    AI
    8 min read

  17. Contrastive Preference Optimization (CPO): Mathematical Foundations, Direct Likelihood Regularization, Out-of-Distribution Error Bounds, and Non-Degenerate Policy Alignment

    Supervised fine-tuning (SFT) and preference alignment represent the standard two-stage training paradigm for post-training large language models. While SFT establishes foundational task adherence by minimizing cross-entropy loss against target references, it suffers from two structural limitations: it cannot surpass the quality ceiling of its training demonstrations, and it lacks an explicit mechanism to penalize near-correct outputs that contain subtle hallucinations or omissions. Direct Prefe

    AI
    11 min read

  18. Post-Training Quantization Toolkits in Production: Comparing AutoAWQ, AutoGPTQ, bitsandbytes, and llm-compressor Architecture, Calibration Profiling, Kernel Execution, and Serving Economics

    Large language model deployment in high-concurrency production environments is governed primarily by memory bandwidth and hardware economics. In standard autoregressive decoding, memory traffic dominates computation: for every token generated, every parameter of the model must be loaded from GPU High Bandwidth Memory (HBM) into SRAM and tensor cores. Serving a 70-billion parameter model in 16-bit precision requires at least 140 GB of VRAM solely for model weights, necessitating multi-GPU tensor

    LLM
    9 min read

  19. Multi-Token Prediction (MTP): Mathematical Foundations, Sequential Latent Stacking, Auxiliary Loss Schedules, and Speculative Inference Acceleration

    Multi-Token Prediction (MTP): Mathematical Foundations, Sequential Latent Stacking, Auxiliary Loss Schedules, and Speculative Inference Acceleration Autoregressive language models have traditionally been trained under a single-token objective: predicting the immediate next token $x_{t+1}$ given the causal context $x_{1:t}$. While this next-token prediction (NTP) paradigm scales predictably with parameter count and dataset volume, it suffers from severe structural limitations. NTP optimizes excl

    AI
    13 min read

  20. Activation Checkpointing and Selective Rematerialization: Mathematical Foundations, Sub-Linear Memory Scaling, Recomputation Trade-Offs, and Megatron-LM Attention Partitioning

    Activation Checkpointing and Selective Rematerialization: Mathematical Foundations, Sub-Linear Memory Scaling, Recomputation Trade-Offs, and Megatron-LM Attention Partitioning Training modern large language models requires orchestrating hundreds of billions of parameters across distributed GPU clusters. While parameter counts and optimizer states are fixed for a given model architecture, the activation memory generated during forward passes scales directly with sequence length, batch size, and

    AI
    10 min read