{"@context": "https://schema.org", "@type": "AnalysisNewsArticle", "inLanguage": "en", "identifier": "ai-research-20260905-0022-607", "headline": "Qwen and Tencent Showcase Translation, Agent, and On-Device Deployment Approaches; Shared-Endpoint Audit Falls Short of Repeatability Thresholds", "description": "Public examples cover multilingual models, on-device translation, research agents, and reusable neural functions generated through training. Task stability, call volume, deployment location, and the costs of updates and retraining jointly affect whether reducing remote requests delivers net benefits in cost, latency, or provider dependence.", "abstract": "Pages and paper entries with record dates up to September 5, 2026, showcase several ways of delivering AI capabilities: Qwen-MT emphasizes multilingual coverage, Hy-MT2 emphasizes on-device translation, and Hyra targets research and engineering tasks. Compilation by training turns natural-language specifications into neural functions that can be saved, versioned, and composed; the resulting functions no longer require the teacher model at runtime[S1][S2][S5]. These examples are useful for comparing task boundaries, execution locations, and dependence on remote services, but they do not establish that the entire industry will shift toward any one approach. A preregistered audit of shared service endpoints recorded repeated-measurement results below its preset thresholds, indicating the need to record request times, service endpoints, replay conditions, and repeated results when comparing models under similar conditions[S6]. Open-source ecosystems should also be examined separately in terms of licensing, weight and code availability, third-party reproduction, and downstream maintenance. Regarding cloud AI infrastructure, the current material supports discussion only of dependence on remote calls and measurement of shared services; it does not support judgments about computing capacity supply, data center investment, or the cloud market landscape.", "datePublished": "2026-09-05T00:22:55+00:00", "dateModified": "2026-09-25T14:22:30.654683+00:00", "author": {"@type": "Person", "name": "Aperture Research Editorial Team", "jobTitle": "Comparative Technology and Industry Research · Research Coder"}, "about": ["Foundation models", "AI agents", "On-device inference", "Open-source ecosystems", "Cloud AI infrastructure", "Model evaluation", "Chinese large models"], "category": "AI industry trends", "authorityScore": 82, "claims": [{"text": "Pages and paper entries with record dates up to September 5, 2026, feature broad-coverage translation models, specialized on-device translation models, agents for research and engineering tasks, foundation models, and neural functions that can be saved and composed[S1][S2][S5]. These examples show that multiple delivery approaches coexist, but they do not establish that the AI industry has already undergone a structural shift. Note that September 5, 2026, is the date attached to the S2 record; the currently visible content does not specify a release date for this Qwen-MT update.", "section": "Compare Five Things Before Comparing Model Scores"}, {"text": "This article uses five comparison criteria: whether capabilities target open-ended or fixed tasks; whether execution takes place through a remote service or in a particular deployment environment; whether runtime operation continues to depend on a large teacher model; whether outputs and evaluations can be repeated under the same conditions; and whether openness extends to weights, licensing, code, and downstream modifications. The last two are comparison criteria, not indications that all the projects concerned have disclosed this information.", "section": "Compare Five Things Before Comparing Model Scores"}, {"text": "The source types are not equivalent: compilation by training, the shared-service audit, and research on operator parallelism on SoCs are covered in arXiv paper entries, while the material on Chinese vendors comes primarily from official pages or webpage link entries[S1][S2][S5][S6][S9]. Accordingly, this article compares publicly described mechanisms, parameters, and product positioning; it does not provide a comprehensive ranking of global vendors. Discussion of cloud AI infrastructure is limited to dependence on remote calls and measurement of shared services, without extending to GPU supply, data center investment, or cloud market share.", "section": "Compare Five Things Before Comparing Model Scores"}, {"text": "The table below includes only information directly supported by the sources. Statements on vendor pages about leadership or performance remain vendor claims; paper entries can describe methods, research questions, or audit results listed in their abstracts, but cannot substitute for validation in commercial deployment.", "section": "Different Approaches Publicly Emphasize Different Bottlenecks"}, {"text": "Compilation by training offers an explicit mechanism: a teacher model first generates task examples from natural-language specifications, then small adapters are trained for use by a compact interpreter; once complete, the resulting neural functions can run without the teacher model[S1]. The outputs of this method can be saved, versioned, and composed[S1]. The currently visible content does not stipulate whether they are ultimately deployed on user devices, edge devices, or in other environments, so “independence from the teacher model at runtime” cannot be directly equated with “local on-device execution.”", "section": "Compilation by Training Moves Some Work Ahead of Runtime"}, {"text": "If the generated functions can replace some existing remote requests, the number of remote calls may decrease, potentially changing cost, latency, and provider dependence. The stability of task boundaries, call volume, deployment location, hardware conditions, and the costs of updating and retraining functions all affect the eventual net benefit. S1 lists recurring costs, latency, and provider dependence as research motivations, but the currently visible content is truncated in the experimental results and does not provide complete cost curves, deployment comparisons, retraining frequency, or end-to-end latency data. It therefore does not directly establish that this method is necessarily cheaper or faster[S1]. Even if the generated functions do not fully replace remote requests, other changes in calling patterns, model selection, or deployment environments may also affect these metrics.", "section": "Compilation by Training Moves Some Work Ahead of Runtime"}, {"text": "Qwen-MT, Hy-MT2, and Hyra cannot be described as using the same mechanism. Qwen-MT publicly emphasizes language coverage and training tokens; Hy-MT2 emphasizes model size, quantization, storage, and mobile deployment; Hyra is positioned as a research and engineering agent capable of recursive self-improvement[S2][S5]. They can be compared within the same framework, but the current sources do not indicate that these products share a common goal of reducing calls to large models.", "section": "Compilation by Training Moves Some Work Ahead of Runtime"}, {"text": "Tencent's page also describes an office workflow that generates editable initial drafts from user materials and requirements[S5]. Such applications can be evaluated by their delivery process—for example, whether they complete a specified deliverable and how much human revision they require. However, S5 provides no corresponding measurements, so it cannot support a ranking of different agents or office systems.", "section": "Compilation by Training Moves Some Work Ahead of Runtime"}, {"text": "The translation products show two different public emphases. The Qwen-MT page emphasizes 92 languages, trillions of multilingual and translation tokens, and reinforcement learning[S2]. Tencent states that Hy-MT2 supports 33 languages while also disclosing on-device metrics, including a 1.8B model, 440MB of storage, and a 1.5-fold improvement in inference speed over the previous version[S5]. The two do not share common test conditions, so language counts or individual device metrics alone cannot establish overall superiority. S2 does not specify a release date for this update, so the information is treated here only as page content visible in the record dated September 5, 2026.", "section": "Chinese Vendor Examples Emphasize Coverage, Deployment, and Task Scope"}, {"text": "Tencent's page also lists Hy3, Hyra, office applications, and on-device translation. The page states that Hy3 uses a 295B/21B MoE architecture and supports a 256K context and three thinking modes, but the listed text does not further explain what parameter-count conventions the 295B and 21B figures respectively represent[S5]. This information confirms that Tencent's public product presentation covers foundation models, research agents, office workflows, and on-device translation. However, S5 does not specify Hy3's execution location or deployment method, nor does it establish that the products have already developed technical or commercial synergies.", "section": "Chinese Vendor Examples Emphasize Coverage, Deployment, and Task Scope"}, {"text": "An open-source ecosystem cannot be represented by performance labels alone. Tencent's page states that Hy-MT2's 7B and 30B-A3B models achieve the best performance among open-source models across multiple types of translation tasks, and that its 1.8B model outperforms some commercial APIs[S5]. These are all vendor performance claims, and the current source does not list independent reproduction results. Whether a sustainable open-source ecosystem has formed also requires separate examination of licensing, weight and code availability, third-party reproduction, and downstream maintenance. S5 does not cover these dimensions, so it cannot support judgments about community governance or capacity for external innovation.", "section": "Chinese Vendor Examples Emphasize Coverage, Deployment, and Task Scope"}, {"text": "Product pages focus on language coverage, parameter counts, storage footprints, or performance on specific tasks[S2][S5]. These metrics can help clarify vendors' chosen product priorities, but release frequency and the number of publicly disclosed metrics do not directly represent industry maturity, nor do they establish that cost pressure drives specialization. Cost, task distribution, and disclosure preferences are all explanations that require further testing.", "section": "Model Scores Also Depend on Measurement Conditions and the Scope of Disclosure"}, {"text": "The shared-service audit shows that measurement conditions themselves need to be included in comparisons. Across 52,988 audited request attempts, a preregistered study recorded a Spearman value of 0.400 for agreement between repeated rankings within the same time window, below the preset threshold of 0.90; the agreement value for next-day replay of byte-for-byte identical requests was 0.78, also below the preset threshold of 0.99[S6]. Within the scope of this shared-endpoint audit, neither repeated rankings within the same window nor next-day replay met the study's preset repeatability criteria. The currently visible results sentence is truncated after 0.78, so the full name of its statistical metric cannot be determined, and it cannot be directly compared statistically with Spearman 0.400.", "section": "Model Scores Also Depend on Measurement Conditions and the Scope of Disclosure"}, {"text": "S6 does not attribute these results to any single established mechanism. Server-side updates, routing, sampling, or other infrastructure changes can only be considered directions for further investigation; this record does not establish any of them as the cause. The results also cannot be generalized into a claim that all cloud models, shared services, or leaderboards are broadly unreliable.", "section": "Model Scores Also Depend on Measurement Conditions and the Scope of Disclosure"}, {"text": "Reasoning text presents another measurement issue. S7 defines a reasoning step's importance as the change it causes in expected reward and asks whether the step's text contains information about its functional role. The paper entry also notes that LLM judges, process reward models, and generative critics rely on the assumption that reasoning text reflects a step's function[S7]. The currently visible content is truncated where the study's definitions are introduced, so it confirms only the research question and the definition of importance. It cannot support a summary of the paper's final experimental results or a general conclusion about the relationship between textual legibility and actual importance.", "section": "Model Scores Also Depend on Measurement Conditions and the Scope of Disclosure"}, {"text": "Foundation models, specialized translation models, research agents, and reusable neural functions can be compared by task boundaries, execution locations, and runtime dependencies[S1][S2][S5]. Compilation by training explicitly proposes allowing generated functions to operate without the teacher model at runtime, but does not stipulate their final deployment location; Hy-MT2 publicly emphasizes on-device quantization and mobile deployment, Qwen-MT emphasizes language coverage, and Hyra targets research and engineering tasks[S1][S2][S5]. These examples show that multiple approaches coexist, but are insufficient to establish that the units of competition or delivery structures across the industry have already differentiated.", "section": "Which Metrics Should Be Used to Test These Approaches Next?"}, {"text": "Comparisons of shared services should record model names, request times, specific endpoints, replay conditions, and repeated-measurement results. S6's audit shows that, within its scope, neither repeated rankings within the same window nor next-day replay of byte-for-byte identical requests met the preregistered thresholds. Because the two results are described differently, and the currently visible content does not provide the full name of the second statistical metric, their magnitudes should not be directly compared, nor should these results be extended into a conclusion about all cloud services[S6]. Evaluation of reasoning steps should distinguish textual clarity from a step's functional importance; S7 provides this research question and the definition of importance, but the currently visible content is insufficient to summarize its final experimental results[S7].", "section": "Which Metrics Should Be Used to Test These Approaches Next?"}, {"text": "Future tracking can focus on complete experimental and deployment data for compilation by training, version and test information for Qwen-MT, and independent reproduction results for the relevant Tencent models[S1][S2][S5]. The DeepSeek record currently shows only the section text “Models & Pricing,” while the MiniMax record shows only the MiniMax Speech 2.8 page title. Both therefore serve only as information entry points and do not contribute to the current substantive assessment[S3][S4].", "section": "Which Metrics Should Be Used to Test These Approaches Next?"}, {"text": "The latest date attached to the records used in this analysis is September 5, 2026, but page collection times or record dates are not equivalent to product release dates[S2][S3][S4]. Source types, publication dates, and test conditions are not uniform; vendor performance statements cannot substitute for independent testing, and results from an individual paper cannot directly explain causal changes across the entire market.", "section": "Which Metrics Should Be Used to Test These Approaches Next?"}], "sections": [{"type": "lead", "heading": "Compare Five Things Before Comparing Model Scores", "paragraphs": ["Pages and paper entries with record dates up to September 5, 2026, feature broad-coverage translation models, specialized on-device translation models, agents for research and engineering tasks, foundation models, and neural functions that can be saved and composed[S1][S2][S5]. These examples show that multiple delivery approaches coexist, but they do not establish that the AI industry has already undergone a structural shift. Note that September 5, 2026, is the date attached to the S2 record; the currently visible content does not specify a release date for this Qwen-MT update.", "This article uses five comparison criteria: whether capabilities target open-ended or fixed tasks; whether execution takes place through a remote service or in a particular deployment environment; whether runtime operation continues to depend on a large teacher model; whether outputs and evaluations can be repeated under the same conditions; and whether openness extends to weights, licensing, code, and downstream modifications. The last two are comparison criteria, not indications that all the projects concerned have disclosed this information.", "The source types are not equivalent: compilation by training, the shared-service audit, and research on operator parallelism on SoCs are covered in arXiv paper entries, while the material on Chinese vendors comes primarily from official pages or webpage link entries[S1][S2][S5][S6][S9]. Accordingly, this article compares publicly described mechanisms, parameters, and product positioning; it does not provide a comprehensive ranking of global vendors. Discussion of cloud AI infrastructure is limited to dependence on remote calls and measurement of shared services, without extending to GPU supply, data center investment, or cloud market share."], "headers": [], "rows": [], "bullets": []}, {"type": "evidence", "heading": "Different Approaches Publicly Emphasize Different Bottlenecks", "paragraphs": ["The table below includes only information directly supported by the sources. Statements on vendor pages about leadership or performance remain vendor claims; paper entries can describe methods, research questions, or audit results listed in their abstracts, but cannot substitute for validation in commercial deployment."], "headers": ["Approach", "Publicly Available Facts", "Potential Optimization Targets", "Current Limitations"], "rows": [["Compilation by training", "A teacher model generates task examples during compilation, after which small adapters are trained for a compact interpreter; the resulting neural functions do not require the teacher model at runtime and can be saved, versioned, and composed[S1]", "The cost, latency, and provider dependence associated with repeated remote calls", "The currently visible content does not stipulate that the resulting functions must run locally on users' devices, nor does it provide complete cost curves, retraining frequency, or commercial deployment data across tasks. It therefore does not establish that the method applies to all text tasks or necessarily improves cost and latency"], ["Broad-coverage translation model", "The official Qwen-MT page states that it is based on Qwen3, uses trillions of multilingual and translation tokens together with reinforcement learning, and supports 92 major official languages and some dialects[S2]", "Language coverage and multilingual translation", "S2 supports these page statements, but the visible body text is truncated in the feature introduction. September 5, 2026, is the record date; the content does not specify a release date for this update, nor does this excerpt establish the model size, storage footprint, or complete test conditions"], ["Specialized on-device translation", "Tencent states that Hy-MT2 supports 33 languages; its 1.8B model occupies 440MB after 1.25-bit quantization, with a 1.5-fold improvement in inference speed over Hy-MT1.5. Tencent also states that the associated mobile app will support local inference[S5]", "Storage footprint, adaptation to mobile chips, and local execution", "The speed improvement, outperformance of commercial APIs, and best performance among open-source models are all vendor claims; the current source does not list independent reproduction results. On-device execution also cannot be directly equated with lower total cost or total latency"], ["Research agent", "Tencent released Hyra-1.0, describing it as an agent capable of recursive self-improvement for performance-oriented research and engineering tasks[S5]", "Research and engineering tasks", "S5 provides no independent data on task completion rates, failure types, or human involvement, so it does not establish real-world task performance"], ["Shared-service evaluation", "A preregistered audit covered 52,988 audited request attempts; the Spearman value for agreement between repeated rankings within the same time window was 0.400, below the preset threshold of 0.90; the agreement value for next-day replay of byte-for-byte identical requests was 0.78, below the preset threshold of 0.99[S6]", "Repeated measurement and service stability", "The currently visible results sentence is truncated after 0.78, so the full name of the statistical metric corresponding to that value cannot be determined; it therefore cannot be treated as the same type of metric as Spearman 0.400. The results apply only within the scope of this audit and cannot automatically be generalized to all cloud models, private services, or local systems"], ["Research on execution across heterogeneous chips", "The Para-Pipe paper's abstract states that traditional pipelining helps improve throughput but does not address the latency requirements of modern neural networks with complex dependencies and extensive operator parallelism; it also discusses the potential to reduce inference latency by executing operators concurrently across multiple processing units[S9]", "Operator parallelism and inference latency on heterogeneous SoCs", "The currently visible content mainly describes the problem setting and potential directions, and is truncated during the subsequent explanation. It therefore cannot support a description of Para-Pipe's complete implementation or an assessment of its actual latency benefits on commercial chips"]], "bullets": []}, {"type": "mechanism", "heading": "Compilation by Training Moves Some Work Ahead of Runtime", "paragraphs": ["Compilation by training offers an explicit mechanism: a teacher model first generates task examples from natural-language specifications, then small adapters are trained for use by a compact interpreter; once complete, the resulting neural functions can run without the teacher model[S1]. The outputs of this method can be saved, versioned, and composed[S1]. The currently visible content does not stipulate whether they are ultimately deployed on user devices, edge devices, or in other environments, so “independence from the teacher model at runtime” cannot be directly equated with “local on-device execution.”", "If the generated functions can replace some existing remote requests, the number of remote calls may decrease, potentially changing cost, latency, and provider dependence. The stability of task boundaries, call volume, deployment location, hardware conditions, and the costs of updating and retraining functions all affect the eventual net benefit. S1 lists recurring costs, latency, and provider dependence as research motivations, but the currently visible content is truncated in the experimental results and does not provide complete cost curves, deployment comparisons, retraining frequency, or end-to-end latency data. It therefore does not directly establish that this method is necessarily cheaper or faster[S1]. Even if the generated functions do not fully replace remote requests, other changes in calling patterns, model selection, or deployment environments may also affect these metrics.", "Qwen-MT, Hy-MT2, and Hyra cannot be described as using the same mechanism. Qwen-MT publicly emphasizes language coverage and training tokens; Hy-MT2 emphasizes model size, quantization, storage, and mobile deployment; Hyra is positioned as a research and engineering agent capable of recursive self-improvement[S2][S5]. They can be compared within the same framework, but the current sources do not indicate that these products share a common goal of reducing calls to large models.", "Tencent's page also describes an office workflow that generates editable initial drafts from user materials and requirements[S5]. Such applications can be evaluated by their delivery process—for example, whether they complete a specified deliverable and how much human revision they require. However, S5 provides no corresponding measurements, so it cannot support a ranking of different agents or office systems."], "headers": [], "rows": [], "bullets": []}, {"type": "case-study", "heading": "Chinese Vendor Examples Emphasize Coverage, Deployment, and Task Scope", "paragraphs": ["The translation products show two different public emphases. The Qwen-MT page emphasizes 92 languages, trillions of multilingual and translation tokens, and reinforcement learning[S2]. Tencent states that Hy-MT2 supports 33 languages while also disclosing on-device metrics, including a 1.8B model, 440MB of storage, and a 1.5-fold improvement in inference speed over the previous version[S5]. The two do not share common test conditions, so language counts or individual device metrics alone cannot establish overall superiority. S2 does not specify a release date for this update, so the information is treated here only as page content visible in the record dated September 5, 2026.", "Tencent's page also lists Hy3, Hyra, office applications, and on-device translation. The page states that Hy3 uses a 295B/21B MoE architecture and supports a 256K context and three thinking modes, but the listed text does not further explain what parameter-count conventions the 295B and 21B figures respectively represent[S5]. This information confirms that Tencent's public product presentation covers foundation models, research agents, office workflows, and on-device translation. However, S5 does not specify Hy3's execution location or deployment method, nor does it establish that the products have already developed technical or commercial synergies.", "An open-source ecosystem cannot be represented by performance labels alone. Tencent's page states that Hy-MT2's 7B and 30B-A3B models achieve the best performance among open-source models across multiple types of translation tasks, and that its 1.8B model outperforms some commercial APIs[S5]. These are all vendor performance claims, and the current source does not list independent reproduction results. Whether a sustainable open-source ecosystem has formed also requires separate examination of licensing, weight and code availability, third-party reproduction, and downstream maintenance. S5 does not cover these dimensions, so it cannot support judgments about community governance or capacity for external innovation."], "headers": [], "rows": [], "bullets": []}, {"type": "counterargument", "heading": "Model Scores Also Depend on Measurement Conditions and the Scope of Disclosure", "paragraphs": ["Product pages focus on language coverage, parameter counts, storage footprints, or performance on specific tasks[S2][S5]. These metrics can help clarify vendors' chosen product priorities, but release frequency and the number of publicly disclosed metrics do not directly represent industry maturity, nor do they establish that cost pressure drives specialization. Cost, task distribution, and disclosure preferences are all explanations that require further testing.", "The shared-service audit shows that measurement conditions themselves need to be included in comparisons. Across 52,988 audited request attempts, a preregistered study recorded a Spearman value of 0.400 for agreement between repeated rankings within the same time window, below the preset threshold of 0.90; the agreement value for next-day replay of byte-for-byte identical requests was 0.78, also below the preset threshold of 0.99[S6]. Within the scope of this shared-endpoint audit, neither repeated rankings within the same window nor next-day replay met the study's preset repeatability criteria. The currently visible results sentence is truncated after 0.78, so the full name of its statistical metric cannot be determined, and it cannot be directly compared statistically with Spearman 0.400.", "S6 does not attribute these results to any single established mechanism. Server-side updates, routing, sampling, or other infrastructure changes can only be considered directions for further investigation; this record does not establish any of them as the cause. The results also cannot be generalized into a claim that all cloud models, shared services, or leaderboards are broadly unreliable.", "Reasoning text presents another measurement issue. S7 defines a reasoning step's importance as the change it causes in expected reward and asks whether the step's text contains information about its functional role. The paper entry also notes that LLM judges, process reward models, and generative critics rely on the assumption that reasoning text reflects a step's function[S7]. The currently visible content is truncated where the study's definitions are introduced, so it confirms only the research question and the definition of importance. It cannot support a summary of the paper's final experimental results or a general conclusion about the relationship between textual legibility and actual importance."], "headers": [], "rows": [], "bullets": []}, {"type": "outlook", "heading": "Which Metrics Should Be Used to Test These Approaches Next?", "paragraphs": [], "headers": [], "rows": [], "bullets": ["Foundation models, specialized translation models, research agents, and reusable neural functions can be compared by task boundaries, execution locations, and runtime dependencies[S1][S2][S5]. Compilation by training explicitly proposes allowing generated functions to operate without the teacher model at runtime, but does not stipulate their final deployment location; Hy-MT2 publicly emphasizes on-device quantization and mobile deployment, Qwen-MT emphasizes language coverage, and Hyra targets research and engineering tasks[S1][S2][S5]. These examples show that multiple approaches coexist, but are insufficient to establish that the units of competition or delivery structures across the industry have already differentiated.", "Comparisons of shared services should record model names, request times, specific endpoints, replay conditions, and repeated-measurement results. S6's audit shows that, within its scope, neither repeated rankings within the same window nor next-day replay of byte-for-byte identical requests met the preregistered thresholds. Because the two results are described differently, and the currently visible content does not provide the full name of the second statistical metric, their magnitudes should not be directly compared, nor should these results be extended into a conclusion about all cloud services[S6]. Evaluation of reasoning steps should distinguish textual clarity from a step's functional importance; S7 provides this research question and the definition of importance, but the currently visible content is insufficient to summarize its final experimental results[S7].", "Future tracking can focus on complete experimental and deployment data for compilation by training, version and test information for Qwen-MT, and independent reproduction results for the relevant Tencent models[S1][S2][S5]. The DeepSeek record currently shows only the section text “Models & Pricing,” while the MiniMax record shows only the MiniMax Speech 2.8 page title. Both therefore serve only as information entry points and do not contribute to the current substantive assessment[S3][S4].", "The latest date attached to the records used in this analysis is September 5, 2026, but page collection times or record dates are not equivalent to product release dates[S2][S3][S4]. Source types, publication dates, and test conditions are not uniform; vendor performance statements cannot substitute for independent testing, and results from an individual paper cannot directly explain causal changes across the entire market."]}], "citations": [{"publisher": "arXiv", "title": "Compile by Training: Turning Natural-Language Specifications into Local Neural Functions", "url": "https://arxiv.org/abs/2609.04199v1", "published_at": "2026-09-03T17:59:49Z", "source_type": "research-paper"}, {"publisher": "Alibaba Qwen", "title": "Next »", "url": "https://qwenlm.github.io/blog/page/2/", "published_at": "2026-09-05", "source_type": "publisher-blog"}, {"publisher": "DeepSeek", "title": "Token & Token Usage", "url": "https://api-docs.deepseek.com/quick_start/token_usage", "published_at": "2026-09-05", "source_type": "publisher-news"}, {"publisher": "MiniMax", "title": "MiniMax Speech 2.8", "url": "https://www.minimaxi.com/news/minimax-speech-28", "published_at": "2026-09-05", "source_type": "publisher-news"}, {"publisher": "Tencent Hunyuan", "title": "Tencent Hunyuan", "url": "https://hunyuan.tencent.com/", "published_at": "2026-09-05", "source_type": "Chinese Large Model Vendor Updates"}, {"publisher": "arXiv", "title": "Clean Engineering, Unstable Measurement: A Preregistered Reliability Failure of Black-Box LLM Observers on Shared Endpoints", "url": "https://arxiv.org/abs/2609.04198v1", "published_at": "2026-09-03T17:59:43Z", "source_type": "research-paper"}, {"publisher": "arXiv", "title": "Legibility is Not Interpretability: Comparing Judged and Actual Importance in Chain-Of-Thought Reasoning", "url": "https://arxiv.org/abs/2609.04194v1", "published_at": "2026-09-03T17:59:08Z", "source_type": "research-paper"}, {"publisher": "arXiv", "title": "Robust PAC Learning of Concurrent Stochastic Games", "url": "https://arxiv.org/abs/2609.04189v1", "published_at": "2026-09-03T17:58:57Z", "source_type": "research-paper"}, {"publisher": "arXiv", "title": "Para-Pipe: Exploiting Hierarchical Operator Parallelism of ML Computational Graphs on SoCs", "url": "https://arxiv.org/abs/2609.04168v1", "published_at": "2026-09-03T17:53:44Z", "source_type": "research-paper"}, {"publisher": "arXiv", "title": "Parameterised graph theory for tensor networks: entanglement rerouting, structural simplification, and agnostic tomography", "url": "https://arxiv.org/abs/2609.04165v1", "published_at": "2026-09-03T17:50:38Z", "source_type": "research-paper"}], "knowledge": {"question": {"en": "How should readers compare Qwen-MT, Hy-MT2 and reusable neural functions, and what does the shared-endpoint audit imply for repeatable model evaluation?", "zh": "应如何比较Qwen-MT、Hy-MT2与可复用神经函数，共享端点审计又对模型评测的可重复性提出了什么要求？"}, "audience": {"en": "Technical evaluators comparing translation products, AI deployment options and shared-service evaluation results.", "zh": "需要比较翻译产品、AI部署方案和共享服务评测结果的技术评估者。"}, "decision": {"en": "Choose comparison criteria and identify missing evidence before selecting a deployment route or trusting a ranking: task boundaries, execution location, runtime dependence on teacher models, repeatability and the scope of openness.", "zh": "在选择部署路线或采信排名前，确定比较标准并识别证据缺口：任务边界、执行位置、运行时对教师模型的依赖、可重复性以及开放范围。"}, "contribution": {"en": "The article distinguishes language coverage, mobile deployment and teacher-independent execution rather than treating them as one technical route. It also uses a preregistered shared-endpoint audit to explain why request times, endpoints, replay conditions and repeated results belong in model comparisons.", "zh": "文章区分了语言覆盖、移动部署和运行时脱离教师模型，而非将其视为同一技术路线；同时借助预注册共享端点审计，说明模型比较为何需要记录请求时间、端点、重放条件和重复结果。"}, "limitation": {"en": "Vendor claims and paper excerpts lack common test conditions and complete deployment evidence. Teacher-independent execution does not establish on-device execution or guaranteed cost and latency savings. The audit neither identifies a cause nor generalizes to all cloud services; its two reported statistics are not directly comparable. The article does not establish an industry-wide transition or cloud-market conclusions.", "zh": "厂商声明与论文摘录缺少共同测试条件和完整部署证据。运行时脱离教师模型并不意味着端侧执行，也不保证成本和延迟下降。审计既未确定原因，也不能推广到所有云端服务；其中两项统计结果不能直接比较。文章不足以确立全行业转向或云市场结论。"}, "nodes": [{"key": "n1", "kind": "concept", "name": {"en": "Compile by training", "zh": "训练式编译"}, "mention": "训练式编译", "id": "concept-90607329e06697946a72"}, {"key": "n2", "kind": "concept", "name": {"en": "Reusable neural functions", "zh": "可复用神经函数"}, "mention": "可复用神经函数", "id": "concept-7ba5f66dcddffcb98958"}, {"key": "n3", "kind": "entity", "name": {"en": "Qwen-MT", "zh": "Qwen-MT"}, "mention": "Qwen-MT", "id": "entity-fef9c90d5eef35b4cdef"}, {"key": "n4", "kind": "entity", "name": {"en": "Tencent Hy-MT2", "zh": "腾讯Hy-MT2"}, "mention": "腾讯Hy-MT2", "id": "entity-c6a409ecc07d1b22278d"}, {"key": "n5", "kind": "concept", "name": {"en": "Shared-endpoint audit", "zh": "共享端点审计"}, "mention": "共享端点审计", "id": "concept-6a9409a9325cdb19dc62"}, {"key": "n6", "kind": "concept", "name": {"en": "Repeated measurement", "zh": "重复测量"}, "mention": "重复测量", "id": "concept-a0d0d915308271da252f"}], "edges": [{"source": "n1", "target": "n2", "relation": "enables", "stance": "reported", "explanation": {"en": "Compile by training uses teacher-generated task examples to train small adapters for a compact interpreter, producing neural functions that can run without the teachers and can be stored, versioned and composed. This mechanism does not specify the final deployment location.", "zh": "训练式编译利用教师模型生成的任务样本，为紧凑解释器训练小型适配器，产出无需教师模型即可运行、并可保存、版本化和组合的神经函数。这一机制未限定最终部署位置。"}, "quote": {"en": "Compile by training provides an explicit mechanism: teacher models first generate task examples from natural-language specifications, then train small adapters for use by a compact interpreter; once completed, the resulting neural functions can run without teacher models [S1]. The products of this method can be stored, versioned and composed [S1].", "zh": "训练式编译提供了一条明确机制：教师模型先根据自然语言规格生成任务样本，再训练供紧凑解释器使用的小型适配器；完成后，生成的神经函数无需教师模型即可运行[S1]。该方法的产物能够保存、版本化和组合[S1]。"}, "citations": [1], "sourceId": "concept-90607329e06697946a72", "targetId": "concept-7ba5f66dcddffcb98958"}, {"source": "n3", "target": "n4", "relation": "compares", "stance": "analysis", "explanation": {"en": "The article contrasts Qwen-MT's emphasis on language coverage and training with Tencent Hy-MT2's disclosed mobile-deployment metrics. These are different public product emphases, not evidence that either product is superior under common conditions; the underlying product figures are vendor claims.", "zh": "文章将Qwen-MT对语言覆盖与训练的强调，同腾讯Hy-MT2披露的移动部署指标进行对照。这体现的是不同的公开产品重点，而非共同条件下的优劣证据；相关产品数字属于厂商口径。"}, "quote": {"en": "The Qwen-MT page emphasizes 92 languages, trillions of multilingual and translation tokens, and reinforcement learning [S2]. Tencent Hy-MT2 claims support for 33 languages, while also disclosing on-device metrics including a 1.8B model, 440MB of storage and an inference speed 1.5 times that of the previous version [S5]. The two lack common test conditions, so overall superiority cannot be judged solely by language counts or individual device metrics.", "zh": "Qwen-MT页面强调92种语言、数万亿规模的多语言与翻译词元以及强化学习[S2]。腾讯Hy-MT2声明支持33种语言，同时披露了1.8B模型、440MB存储和相对上一版本1.5倍的推理速度等端侧指标[S5]。两者没有共同测试条件，不能只按语言数量或单项设备指标判断整体优劣。"}, "citations": [2, 5], "sourceId": "entity-fef9c90d5eef35b4cdef", "targetId": "entity-c6a409ecc07d1b22278d"}, {"source": "n5", "target": "n6", "relation": "measures", "stance": "reported", "explanation": {"en": "Within this audit, both same-window ranking repetition and next-day replay fell below their preregistered thresholds. The second statistic's full name is unavailable in the visible excerpt, so it must not be treated as the same measure as Spearman correlation. The findings do not identify the mechanism responsible.", "zh": "在此次审计范围内，同窗口排名重复和次日重放均低于预注册门槛。可见摘录未提供第二项统计指标的完整名称，因此不能将其视为与Spearman相关系数相同的指标；这些结果也未确定造成差异的机制。"}, "quote": {"en": "A preregistered study recorded, across 52,988 audited request attempts, a Spearman value of 0.400 for agreement between repeated rankings within the same time window, below the preset 0.90; the agreement value for next-day replays of byte-identical requests was 0.78, also below the preset 0.99 [S6]. Within the scope of this shared-endpoint audit, neither same-window repeated rankings nor next-day replay met the study's preset repeatability standards.", "zh": "一项预注册研究在52,988次经审计的请求尝试中记录到，同一时间窗口重复排名的一致性Spearman值为0.400，低于预设的0.90；逐字节相同请求的次日重放agreement值为0.78，也低于预设的0.99[S6]。在此次共享端点审计范围内，同窗口重复排名和次日重放都没有达到研究预设的重复标准。"}, "citations": [6], "sourceId": "concept-6a9409a9325cdb19dc62", "targetId": "concept-a0d0d915308271da252f"}]}, "knowledgeMap": "/knowledge", "license": {"model": "open", "price": 0.002, "currency": "USDC", "paymentProtocol": "x402", "network": "eip155:84532", "payTo": "0x23a885262A89c32eB582f4439184769dA22c467c", "variants": {"A": "/agent/v1/articles/ai-research-20260905-0022-607?lang=en", "B": "/agent/v1/articles/ai-research-20260905-0022-607/paid?lang=en"}}, "accessVariant": "A", "contentPolicy": {"citationAllowed": true, "attributionRequired": true, "trainingUse": "contact publisher"}}