#Multimodal

Browser Use releases open-source browser automation benchmark BU Bench V1, Claude Fable-5 scores up to 80.0%

Browser Use released the open-source browser automation benchmark BU Bench V1, containing 100 hardcore browser tasks, Claude Fable-5 scored the highest at 80.0%, leading GPT 5.5 by 12 points; Browser Use Cloud (bu-ultra) scored 78.0%, with a speed of approximately 14 tasks per hour. The benchmark is based on real-world website tasks, scored using LLM evaluators, and is available on GitHub.

INTELLECT-3: A 100B+ parameter expert mixed model trained using large-scale reinforcement learning

INTELLECT-3 is a 106B parameter expert mixed model, trained on the GLM 4.5 Air base model using SFT and RL, achieving optimal performance at its scale on math, code, science, and reasoning benchmarks, surpassing many larger front-line models.

Environments Hub: A community-driven RL environment Open AGI center

Environments Hub released, an open community platform for extending RL to Open AGI; supports development, sharing, and reusing RL environments, reducing friction in building, sharing, and reusing environments, promoting open-source AGI progress; now open to everyone, providing RL, Reinforcement Fine-tuning (RFT), computing, and reasoning infrastructure.

Prime Intellect previews its distributed reasoning stack, building a reasoning engine for the public internet.

Prime Intellect previews its distributed reasoning stack, designed for consumer-grade GPUs and public internet 100ms latency, and releases three open-source code repositories PRIME-IROH, PRIME-VLLM, PRIME-PIPELINE to achieve latency-aware pipeline parallelism, aiming to achieve the openness and democratization of AGI through consumer-grade hardware and network connections.

Anthropic research on how to trace the thinking process of the large language model Claude.

Anthropic has published two new papers, constructing an AI microscope to trace the internal mechanisms of the large language model Claude. The research found that Claude has a shared conceptual space when processing multilingual sentences, can plan the output of multiple words in advance, and provide answers that meet user expectations rather than logical reasoning when needed. These findings help understand AI capabilities and ensure its reliability, but current methods still have limitations and need improvement to handle more complex models.

Enhance Dify RAG with InfraNodus: Expand Your LLM Context

InfraNodus enhances Dify RAG by generating an overview of the knowledge base, helping it better understand context and improve the quality of answers to open-ended questions. InfraNodus extracts metadata such as topics, relationships, and blind spots, enhancing RAG prompts to enable LLMs to cover a broader range of important issues. InfraNodus can process data like PDFs and text, and supports API-driven automated insight extraction, significantly improving the structure and accuracy of answers after integration with Dify.

DeepSeek releases SYNTHETIC-1: a large-scale distributed synthetic reasoning dataset

Dify has released the Plugins plugin system to modularize component extensions to AI application capabilities.

Dify has launched the Plugins plugin system into Beta, extending AI application functionality through plug-and-play components. Developers can obtain ready-made plugins or build their own on the Dify Marketplace, supporting capabilities such as image processing, audio analysis, text generation, data analysis, logical reasoning, and external integration. The system adopts a decoupled architecture, including five types of components: models, tools, strategies, extensions, and packages, achieving bidirectional communication with Dify's core functionality through custom endpoints.

How Dify Enhances GPT-Researcher Through Parallel and Advanced Iterative Features

Dify Workflow accelerates GPT-Researcher's research tasks through parallel mode and advanced iterative features, breaking down the problem into 4-6 independent sub-problems, parallel extracting topics to form subheadings, calling the Tavily API to search and analyze results, and finally integrating them into a structured report, significantly shortening research time and isolating error handling.

How Dify.AI empowers this global consumer electronics leader to integrate generative AI

The global consumer electronics leader serving over 100 million users has integrated generative AI into its operations through Dify.AI, bridging the gap between business and technology. Dify.AI's low-code platform allows novice users to quickly utilize mainstream models, expert users to independently customize solutions, and developers to enhance efficiency through component-based workflows. More than 200 AI applications have been created within a month, with 60 in use, reducing user voice analysis time from 8 hours to 3 hours.

Anthropic Research: Evaluating the Effectiveness of Feature-Guided Mitigation of Social Bias

Anthropic conducted guiding experiments on 29 social bias-related features of Claude 3 Sonnet, finding that moderate guidance can reduce bias without compromising model capabilities, but excessive guidance may lead to a decline in model performance or even make it unusable; some guidance can precisely influence specific domain evaluation results, but the effects of features are difficult to predict solely based on their triggered contexts, leading to off-target effects; one neutral behavioral feature can effectively reduce bias across nine dimensions without significantly affecting capabilities.

Dify v0.10.0 released: Supporting file upload functionality.

Dify v0.10.0 released, adding file upload functionality, supporting document, audio, and video files, enabling applications such as Q&A and summary generation, and allowing the creation of AI-driven podcasts through file upload and LLM nodes.

Anthropic research on how to measure the persuasiveness of large models.

Anthropic developed a method to measure the persuasiveness of large models, comparing Claude 1, 2, and 3 generations, as well as compact and cutting-edge models, and found that the persuasiveness of each generation of models has shown an upward trend, with the latest model Claude 3 Opus having persuasiveness that is not significantly different from that of humans.

Dify releases new architecture Dify Beehive, enhancing flexibility and scalability.

Dify refactors its architecture into a modular design, launching the Model Runtime service module to support models with 175B parameters, achieving a unified interface and simplified YAML configuration, dynamically adding models, improving development efficiency and compatibility.

Dify.AI integrates the Jina AI embedding model to enhance the performance of RAG applications.

Dify.AI integrates the Jina Embeddings model into its platform, which supports a context length of 8,192 tokens, surpassing most open-source models that only support 512 tokens.

Dify.AI v0.3.29 Released: Supports GPT-4 Turbo and Vision Models

Dify.AI v0.3.29 version adds support for OpenAI GPT-4 Turbo (128K context) and GPT-4 Vision models, and introduces external data API tools and sensitive content review features.

Dify v0.3.13 released: Easily integrate open-source large models like Llama2 and ChatGLM

Dify v0.3.13 released, integrating open-source large models like Llama2 and ChatGLM, supporting models from Hugging Face and Replicate platforms, allowing users to easily access and quickly build AI applications via API keys. The platform supports rapid switching between models, enabling users to compare performance by clicking the model selector, lowering the barrier for model selection and tuning.