# Open Source的所有文章

在 HTX 新闻中心浏览与「Open Source」相关的最新资讯与深度分析。潘盖市场趋势、项目动态、技术进展及监管政策，提供权威的加密行业洞察。

China's No.1, Closing in on OpenAI, Mysterious "Sweeping Monk" Rises to Top Seven Globally

A mysterious Chinese AI project named "MopMonk" (meaning "Sweeping Monk") has achieved a top-ranking result on the globally recognized CyberGym cybersecurity benchmark. With a 73.1% success rate, it ranks seventh worldwide and first among Chinese entries, performing closely behind OpenAI. The significance lies in the benchmark itself. CyberGym, created by UC Berkeley, is considered a premier "Olympics" for AI security. It tests models on over 1500 real-world software vulnerabilities, requiring them to not just identify but actually generate working exploits (PoCs) in a complex, offline environment. This moves beyond simple knowledge to testing an AI's practical "execution" capabilities. MopMonk's approach is notable. It uses the open-source MiniMax M3 model from Shanghai as its powerful reasoning "brain," leveraging its strong coding skills and long context window. However, the key to its performance is a custom-built, multi-agent security framework—its "Harness." This system uses structured "vulnerability memory" to efficiently guide the search for exploits, allowing multiple agents to explore in parallel while sharing lessons learned from failures. This engineering layer effectively translates the model's intelligence into actionable, iterative testing steps. The project remains highly secretive, with no official website or team information, embodying the "dark horse" spirit of its literary namesake. Its success highlights a potential industry shift: beyond simply scaling model size, the engineering of specialized agent systems (the Harness) is becoming a critical differentiator for real-world AI application performance, especially in complex domains like cybersecurity.

marsbit刚刚

China's No.1, Closing in on OpenAI, Mysterious "Sweeping Monk" Rises to Top Seven Globally

marsbit刚刚

Deforming the Transformer, LLMs Become Smarter

A new research paper proposes "Tapered Language Models (TLMs)," a method that improves large language model performance without adding any parameters. It challenges the standard Transformer design where each layer has the same number of parameters ("feed-forward network" width). Building on evidence that layers are not equally important—earlier layers handle foundational information like grammar, while later layers often reinforce existing judgments—the researchers suggest reallocating model capacity from later to earlier layers. The core idea is to make the layer width taper off monotonically from start to end, keeping total parameters and compute constant. Experiments compared linear, cosine, and sigmoid tapering curves on a 440M parameter model. The cosine curve (e.g., starting width 1.5x baseline, ending 0.5x) achieved the best result, reducing perplexity by 1.84 points compared to the uniform baseline—a significant gain at zero cost. This finding proved robust across four different model architectures (including gated attention and memory-augmented models) and at larger scales (760M and 1.3B parameters), consistently improving performance on commonsense reasoning and language modeling tasks without harming long-context retrieval ability. The work highlights a long-overlooked design dimension: optimal parameter allocation across depth. It offers a "free lever" for efficiency, potentially applicable beyond language models to vision Transformers and diffusion models. The study was conducted by researchers from Mila, Cornell University, and the University of Montreal.

marsbit19小时前

Deforming the Transformer, LLMs Become Smarter

marsbit19小时前

I Only Trust yuxinlu1 in the Hugging Face Model TOP Rankings Now

An independent developer has unexpectedly risen to the top of Hugging Face's Trending models chart, surpassing major tech companies. The developer, Yuxin Lu (yuxinlu1), has two 12B GGUF models (Coder and Agentic versions) based on Gemma-4-12B on the list, with downloads exceeding 200k and 500k respectively. His models distill coding and reasoning capabilities from sources like Fable 5 into a locally runnable format, requiring as little as 4.5GB of VRAM. Lu, a graduate student in AI, developed these models as a personal, self-funded project for skill development. He emphasizes data quality over quantity, using around 10,000 verified examples. The models' popularity stems from offering privacy and a free, offline alternative for coding assistance and agentic tasks, filling a niche not prioritized by larger companies focused on broader goals. Lu's other projects include general-purpose distillation models and Chinese web novel LoRAs. He advises fellow developers to prioritize honesty about model capabilities and persistence. He views his success not as defeating major players, but as focusing deeply on solving specific user needs with genuine effort. The models are best run using llama.cpp.

marsbit前天 01:52

I Only Trust yuxinlu1 in the Hugging Face Model TOP Rankings Now

marsbit前天 01:52

US Large Models Moving Towards Closure, In the Name of Security

Summary: The U.S. government has intervened to regulate the release of advanced AI models, citing national security concerns. This analysis details the events surrounding the controlled rollout of Anthropic's Mythos 5 and Fable 5 models and OpenAI's restricted GPT-5.6 release, following a Trump administration executive order. It questions the technical basis for the government's actions, noting that internal company safety assessments did not deem the models an unacceptable risk. The article draws parallels to the 1990s "Crypto Wars," where similar attempts to control encryption technology ultimately failed, harming U.S. competitiveness. Experts warn that opaque, non-technical regulatory processes could stifle innovation, concentrate power, and undermine the economic logic of the AI industry, especially as Chinese models embrace open-source development. The piece concludes that this marks a potential turning point toward a pre-approval era for U.S. AI, creating uncertainty for developers and potentially slowing the broad societal diffusion essential for a technology's success.

链捕手前天 15:21

US Large Models Moving Towards Closure, In the Name of Security

链捕手前天 15:21

Dawn Song, The First Lady of Computer Security, Joins Meta

Computer security and AI safety pioneer Dawn Song has joined Meta's Superintelligence Labs as Vice President of AI Research, reporting to MSL head Nat Friedman. Song, a professor at UC Berkeley's EECS department and a MacArthur Fellow, is renowned for her foundational work in dynamic taint analysis and her leadership in adversarial machine learning and AI agent security. Her lab is considered a top training ground in computer security. Song is also a founder of Oasis Labs and Virtue AI, a company focused on enterprise AI safety infrastructure. Virtue AI co-founders Bo Li and Sanmi Koyejo, along with other team members, are also joining Meta, a move seen as strengthening Meta's safety measures for AI agents amid growing industry concerns. Separately, Denny Zhou, founder of Google's Gemini Reasoning Team and a key figure in advancing large language model reasoning techniques like Chain-of-Thought, reportedly joined Meta several months ago. These high-profile hires come as Meta seeks to deploy AI across its products while assuring regulators and the public of its models' robustness against misuse.

marsbit06/26 08:12

Dawn Song, The First Lady of Computer Security, Joins Meta

marsbit06/26 08:12

NVIDIA's New Open-Source MoE: One Line of Import, Fine-Tuning Accelerated by 3.7x

NVIDIA has open-sourced NeMo AutoModel, a tool designed to significantly accelerate the fine-tuning of Mixture-of-Experts (MoE) large language models. By adding just one import line to existing code based on Hugging Face Transformers v5, users can achieve a 3.4x to 3.7x increase in training throughput and reduce GPU memory usage by 29% to 32% without altering their API. The key innovations include Expert Parallelism (EP) to distribute expert weights across GPUs, lowering memory pressure; DeepEP to fuse computation and communication; and TransformerEngine kernels for accelerated core operations. Benchmarks on models like Qwen3-30B-A3B show training throughput per GPU jumping from 3075 to 11340 tokens per second. The solution also enables the fine-tuning of very large models, such as the 550B parameter Nemotron 3 Ultra, which would exceed memory limits with the standard Transformers v5. Code and benchmarks are available on GitHub.

marsbit06/26 07:29

NVIDIA's New Open-Source MoE: One Line of Import, Fine-Tuning Accelerated by 3.7x

marsbit06/26 07:29

Why Does 'AGI Godfather' Ben Goertzel Believe the Future of AI Relies on Blockchain?

Ben Goertzel, known as the "AGI Godfather," argues that the future of Artificial General Intelligence (AGI) must be built on blockchain to prevent its control by a few corporations or venture capital firms. He believes the core AGI code should be free and open-source, but that this alone is insufficient without a decentralized infrastructure to run it affordably. His blockchain project, SingularityNET, and the broader Artificial Superintelligence Alliance aim to create a user-owned, decentralized network for hosting and deploying AGI, contrasting with the closed models of companies like OpenAI and Anthropic. Goertzel criticizes the shift of other labs from open to closed development. He argues that while a closed path is simpler, an open, decentralized model—akin to Linux and the internet—is both possible and ultimately better for humanity. He envisions an "Agent economy" where individuals orchestrate teams of AI agents to perform tasks, including transactions, on an open network rather than corporate clouds. While his current model relies on cryptocurrency, plans include offering paid AI services to businesses with the decentralized blockchain as the backend. Goertzel predicts human-level AGI could arrive by 2029 and warns that a gap in understanding and access to AGI could drastically worsen inequality. The first test of his decentralized approach will be the upcoming release of the Agent Omega Claw.

Foresight News06/22 12:10

Why Does 'AGI Godfather' Ben Goertzel Believe the Future of AI Relies on Blockchain?

Foresight News06/22 12:10

Open Systems Will Ultimately Prevail: Why Ethereum Is the Next Linux?

The article "Open Systems Will Ultimately Prevail: Why Ethereum Is the Next Linux?" argues that Ethereum, like Linux before it, will triumph over closed, proprietary systems in finance due to its open, permissionless, and credibly neutral nature. It draws a historical parallel: just as the open internet defeated corporate private networks and Linux outcompeted proprietary Unix systems, open financial infrastructure like Ethereum will surpass private blockchains. The core advantage lies in the "bazaar" development model (as described in Eric Raymond's "The Cathedral and the Bazaar"), where decentralized, permissionless innovation by a global community of developers outpaces the controlled "cathedral" approach of centralized entities. This model fosters rapid innovation, as seen with Ethereum standards like ERC-20 and applications like Uniswap, which were built without needing permission. Ethereum's key, irreplicable strength is its credible neutrality: transparent, equally applicable, immutable rules that allow anyone to participate. This ensures sovereign independence, meaning no single entity (company, government) can control or change its core rules—a critical feature for global financial infrastructure. In contrast, private blockchains and consortium chains (like SWIFT or various bank-led projects) suffer from platform risk, central control, and an inability to attract broad developer ecosystems, leading to frequent failures. The article notes that major institutions (e.g., BlackRock, JPMorgan, Coinbase, Robinhood) are already building on Ethereum or its Layer 2 networks, recognizing its security, developer ecosystem, and network effects. While critics argue finance requires accountable, controlled systems, the response is that compliance (KYC, regulations) can be built at the application layer on top of a neutral settlement layer like Ethereum, just as secure commerce was built on the open internet via HTTPS. Ultimately, the thesis is that attempting to build walled-garden, proprietary financial networks is a flawed strategy that stifles innovation. The winning approach is to build applications on top of open, credibly neutral infrastructure like Ethereum, which is poised to become the foundational settlement layer for global finance.

Foresight News06/22 10:28

Chips, Open-Source Models, and $50 Trillion: Joe Tsai Reassesses Alibaba Once Again

Alibaba Executive Chairman Joe Tsai recently outlined the company's comprehensive AI strategy in a public discussion. He believes AI represents a massive opportunity, estimating its potential economic impact at up to $50 trillion, stemming from the automation of human intelligence and productivity. Tsai detailed Alibaba's four-layer investment approach across the AI stack: starting from the chip level, moving to cloud infrastructure (Alibaba Cloud), then the model layer with its open-source Qwen model, and finally applications within its vast digital ecosystem (e-commerce, logistics, etc.). The company avoids the energy layer due to China's efficient infrastructure. This broad strategy is designed to ensure Alibaba captures value regardless of where it ultimately concentrates in the AI value chain. He dismissed concerns about an AI investment bubble, pointing to the enormous $50 trillion opportunity. While acknowledging U.S. cloud giants' higher capital expenditure, he argued Chinese firms, including Alibaba (funded by its cash-generative e-commerce core), need to invest more in AI infrastructure. A key theme was technological sovereignty. Tsai positioned open-source models like Qwen as a solution for companies, especially in Europe, seeking independence from proprietary U.S. models and greater data privacy control. He contrasted this with the trend of U.S. giants keeping their models closed-source. Tsai highlighted Alibaba's collaborations with European manufacturers like Bosch and Siemens, using AI for design and quality control. He concluded with an optimistic vision of AI agents enhancing productivity, ultimately freeing up human time for leisure, family, and experiences like live entertainment.

marsbit06/22 07:51

Chips, Open-Source Models, and $50 Trillion: Joe Tsai Reassesses Alibaba Once Again

marsbit06/22 07:51

Ethereum Is Retracing the Path of the Internet and Linux: No One Yields, and the Neutral Party Ultimately Prevails

This article argues that Ethereum is following the historical path of open, neutral systems like the Internet and Linux, which eventually triumphed over proprietary, centrally-controlled alternatives. Major financial institutions like JPMorgan, Stripe, and Circle are building their own proprietary blockchains or networks (e.g., Tempo, Arc), but will never agree to build on a competitor's controlled infrastructure. This creates the perfect opportunity for Ethereum as the only neutral, credibly neutral settlement layer that no single entity controls. The piece draws parallels to the 1990s, when experts like Bill Gates predicted proprietary networks (from Microsoft, Oracle) would win over the open Internet, and when Sun Microsystems' Unix lost to the open-source "bazaar" development model of Linux. This model, described in Eric Raymond's "The Cathedral and the Bazaar," thrives on permissionless innovation where countless contributors improve the system, outpacing any centralized competitor. Ethereum embodies this through its decentralized development, broad validator distribution, and credible neutrality—rules that are transparent, equally applied, hard to change, and open to all. This has attracted over a million developers and major institutions like Coinbase, BlackRock, and JPMorgan, who choose Ethereum for its security, ecosystem, and sovereignty (the inability of any single party to change the rules). While proprietary chains offer initial speed and control, they inherit the downsides of both centralization and decentralization without the long-term innovation benefits. The article concludes that, just as open systems historically win, Ethereum is poised to become the foundational, neutral settlement layer for global finance.

marsbit06/22 02:51

Ethereum Is Retracing the Path of the Internet and Linux: No One Yields, and the Neutral Party Ultimately Prevails