{"@context":"https://schema.org","@type":"NewsArticle","generatedAt":"2026-07-23T09:05:06.107Z","headline":"DharmaOCR 在巴西葡萄牙语 OCR 任务上以 0.925 分超越 Mistral OCR4 与 Unlimited-OCR","description":"DharmaOCR 在针对巴西葡萄牙语的基准测试中取得 0.925 分，而 Mistral OCR4 和 Unlimited-OCR 分别仅获 0.798 和 0.7587 分。该模型通过两阶段训练--监督微调结合直接偏好优化（DPO）--将全部参数集中于单一语言领域，从而在提取质量和稳定性上超越参数规模更大的多语言模型。这一优势源于领域专精而非架构大小，且在新模型发布后依然成立。","url":"https://www.aioga.com/news/cmrni9h7w00c7bii5am7uma4i/","mainEntityOfPage":"https://www.aioga.com/news/cmrni9h7w00c7bii5am7uma4i/","datePublished":"2026-07-16T11:49:48.000Z","dateModified":"2026-07-16T11:49:48.000Z","inLanguage":"zh-CN","publisher":{"@type":"NewsMediaOrganization","name":"Aioga","url":"https://www.aioga.com"},"citation":["https://huggingface.co/blog/Dharma-AI/newer-models-same-advantages","https://aihot.virxact.com/items/cmrni9h7w00c7bii5am7uma4i"],"canonicalUrl":"https://www.aioga.com/news/cmrni9h7w00c7bii5am7uma4i/","directAnswer":{"@type":"Answer","text":"Aioga 编辑摘要：DharmaOCR 在针对巴西葡萄牙语的基准测试中取得 0.925 分，而 Mistral OCR4 和 Unlimited-OCR 分别仅获 0.798 和 0.7587 分。 Aioga 将其归入「AI资讯」方向，重点关注它对真实使用和行业竞争的影响。","url":"https://www.aioga.com/news/cmrni9h7w00c7bii5am7uma4i/","dateCreated":"2026-07-16T11:49:48.000Z","author":{"@type":"Organization","@id":"https://www.aioga.com/authors/aioga-editorial/#editorial-team","name":"Aioga Editorial Team","url":"https://www.aioga.com/authors/aioga-editorial/"}},"evidence":[{"@type":"CreativeWork","name":"huggingface.co source article","url":"https://huggingface.co/blog/Dharma-AI/newer-models-same-advantages","datePublished":"2026-07-16T11:49:48.000Z","provider":{"@type":"Organization","name":"huggingface.co","url":"https://huggingface.co/blog/Dharma-AI/newer-models-same-advantages"}},{"@type":"CreativeWork","name":"AIHot archive record","url":"https://aihot.virxact.com/items/cmrni9h7w00c7bii5am7uma4i","datePublished":"2026-07-16T11:49:48.000Z","provider":{"@type":"Organization","name":"AIHot","url":"https://aihot.virxact.com/items/cmrni9h7w00c7bii5am7uma4i"}}],"aggregationSource":"Hugging Face：Blog（RSS）","originalPublisher":{"name":"huggingface.co","url":"https://huggingface.co/blog/Dharma-AI/newer-models-same-advantages"},"article":{"id":"cmrni9h7w00c7bii5am7uma4i","slug":"cmrni9h7w00c7bii5am7uma4i","url":"https://www.aioga.com/news/cmrni9h7w00c7bii5am7uma4i/","title":"DharmaOCR 在巴西葡萄牙语 OCR 任务上以 0.925 分超越 Mistral OCR4 与 Unlimited-OCR","title_en":"Newer Models， Same Advantage","summary":"DharmaOCR 在针对巴西葡萄牙语的基准测试中取得 0.925 分，而 Mistral OCR4 和 Unlimited-OCR 分别仅获 0.798 和 0.7587 分。该模型通过两阶段训练--监督微调结合直接偏好优化（DPO）--将全部参数集中于单一语言领域，从而在提取质量和稳定性上超越参数规模更大的多语言模型。这一优势源于领域专精而非架构大小，且在新模型发布后依然成立。","source":"Hugging Face：Blog（RSS）","sourceUrl":"https://huggingface.co/blog/Dharma-AI/newer-models-same-advantages","aiHotUrl":"https://aihot.virxact.com/items/cmrni9h7w00c7bii5am7uma4i","publishedAt":"2026-07-16T11:49:48.000Z","category":"AI资讯","score":0,"selected":false,"articleBody":["Three months ago, we published a paper on DharmaOCR：https://arxiv.org/abs/2604.14314 and open-sourced one of the models：https://huggingface.co/Dharma-AI/Dharma-OCR-LITE. The objective was specific: optical character recognition engineered for Brazilian Portuguese.","The training pipeline was built in two stages. The first was a supervised fine-tuning step, drawing on a broad collection of Portuguese-language files from different sources, formats, and levels of complexity. This stage aligned the model's weights to the specific vocabulary, syntax, and document structures of Brazilian Portuguese — concentrating representational capacity on the target language rather than distributing it across a broader multilingual space. The second stage applied Direct Preference Optimization: rather than training only on correct transcriptions, the model learned from comparative preference data between competing outputs, teaching it to consistently select the better extraction at inference time. This stage addressed a different problem: not accuracy, but stability. By suppressing the failure modes that cause generative models to produce repetitive or incoherent output, DPO reduced inference time and cost, and materially improved the reliability of what the model delivered in production.","The combined result was a model that achieved the highest extraction quality score with the lowest degeneration rate on a Portuguese-focused benchmark. Both stages were necessary. The fine-tuning stage built domain competency; the DPO stage ensured that competency held under the conditions where models tend to fail.","OCR models have been moving quickly. But the gaps that originally motivated DharmaOCR's design (in extraction quality on complex documents and in model stability under production conditions) have not closed. They have, if anything, become more instructive as the field has changed.","The proliferation of multimodal generative models made language model-based OCR widely accessible, and the wave of fine-tuned OCR variants that followed reflects how fast that adoption has moved. That proliferation has not, however, changed the fundamental character of the technology. Every OCR system built on a generative model is probabilistic. Transcription errors are an inherent variable of this probabilistic technology. What differentiates models is how many errors they make and of what kind. That is determined by two things: the structure of the model (its architecture and parameter count) and how those parameters were trained for the task.","Architecture and parameter count establish the ceiling on what a model can learn. Training determines how that capacity is allocated.","This distinction is where specialization becomes a structural question rather than a design preference. When a model is trained on a restricted domain — a single language, a bounded document type, a specific task — All of its parameters are dedicated to that specific task. When a model is trained to cover a broader range of domains — a multilingual model handling N languages, for instance — those same parameters must be distributed across all of them. The distribution is not linear: the neuron superposition principle means individual parameters can encode multiple features simultaneously. But the division is real, and its consequences are real. A model covering more ground commits less to any given part of it.","DharmaOCR was trained to accept that constraint in reverse. The model is not designed to be the best option for other languages, and was never intended to be. In exchange, every parameter available to the network could be oriented toward the specific vocabulary, morphology, and orthographic patterns of Brazilian Portuguese — the most directed possible use of the model's resources for that domain.","That concentration is the structural basis of an inherent advantage over multilingual and broader-domain models. The advantage does not depend on having a larger architecture or a more sophisticated training procedure than competitors use — new architectures and new training techniques improve what any model can do. It depends on where those resources are directed: at one domain rather than spread across many.","Three months later, newer models have arrived. Whether the case for specialization holds when those models are newer and more capable is a different question.","Three months after the DharmaOCR paper appeared, two new OCR models attracted significant attention from the research community: Mistral OCR4 and Unlimited-OCR. Both represent genuine technical advances — new training techniques, new datasets, and strong results across multiple languages on a range of benchmark evaluations. They are the kind of models that raise the competitive standard for what OCR systems are expected to deliver.","When we ran both against the DharmaOCR benchmark — an evaluation designed exclusively around Portuguese — the results were conclusive.","DharmaOCR scored 0.925. Mistral OCR4 scored 0.798. Unlimited-OCR scored 0.7587.","The difference is significant. Mistral OCR4 falls approximately 13 points below DharmaOCR; Unlimited-OCR falls more than 16 points below. Both were released after our model, both backed by substantial research resources. On a task where DharmaOCR's fundamental design decision was to concentrate entirely on Portuguese, the specialization advantage is measurable and significant.","The benchmark is the central finding. What follows illustrates why the gap takes the specific shape it does.","Processing non-trivial Portuguese documents reveals precisely where multilingual models tend to break. ENEM essays (Brazil's national high school examination) combine handwritten text with vocabulary, proper nouns, and cultural references that are specific to Brazilian Portuguese. They are exactly the kind of documents where language-specific training produces a return.","：https://cdn-uploads.huggingface.co/production/uploads/67b22efa9b36260fe286e99f/XVbylBZ_TyY98EvfkO9pT.png","Figure 1: ENEM essay manuscript used in benchmark evaluation with outputs of the each model. Misreads marked in red.","Mistral OCR4, evaluated on documents of this kind, transcribed the name Chico Buarque (one of Brazil's most widely recognized musicians and poets) as \"Chico Barque.\" Unlimited-OCR rendered the same name as \"chico bique.\" Confronted with the phrase \"O Brasil não exclui, assimila\" (\"Brazil does not exclude, it assimilates\", a Chico Buarque quotation embedded in the same document) Unlimited-OCR returned: \"a dose de chico bique, 'o Brasil no exclu, eliminila.'","These are not random errors. A model with insufficient exposure to Brazilian Portuguese does not fail arbitrarily — it fails at precisely the vocabulary and proper nouns that distinguish Brazilian Portuguese from the broader multilingual corpus. Chico Buarque is not an obscure reference; the name is nationally recognized. Its systematic corruption across outputs is not an edge case. It is a diagnostic: evidence of where the model's training did not go.","DharmaOCR, evaluated on the same documents, handles these cases correctly. The reason is direct: the model's training was concentrated on this linguistic space, orienting its resources toward the vocabulary and proper noun distributions that characterize Brazilian Portuguese rather than spreading them across many languages at once.","The examples illustrate the benchmark rather than replace it. The benchmark establishes the magnitude of the gap; the examples show why it is concentrated in language-specific recognition rather than in general capability.","Extraction accuracy, however, is only one dimension of production performance. Stability under visual difficulty is another — and operationally, it is the more consequential one to fail on.","When a generative model encounters a document it cannot clearly resolve — small fonts, degraded scan quality, dense handwriting — it faces uncertainty in its input signal. Models trained primarily on next-token prediction objectives face a specific vulnerability here: when the visual signal becomes ambiguous, the model can continue generating from prior learned patterns rather than from the source document. The result is text degeneration — output that is repetitive, incoherent, and semantically disconnected from the page.","Presented with a document with small fonts, Mistral OCR4 produces output with no connection to what is written.","：https://cdn-uploads.huggingface.co/production/uploads/69d815b52c6db28cfdfdd422/3YRHeWDsJ8twDlf7ArbZr.png","：https://cdn-uploads.huggingface.co/production/uploads/67b22efa9b36260fe286e99f/yBwDuIrqnY0xQ8RLY5Gbe.png","Figure 3: DharmaOCR output and Mistral OCR4 degenerated output","This is not a low-quality transcription of the source. It is a failure of an entirely different category.","The operational consequence is distinct from that of a transcription error. An incorrect transcription is wrong in a recoverable way — it stands in a relationship to the source document, and the error can in principle be identified and corrected. Degenerated output has no such relationship. It cannot be corrected because there is nothing to correct toward. For downstream processes that depend on structured OCR output — document classification, information extraction, compliance workflows — degenerated output is not inaccurate data. It is structurally unusable data. The efficiency that automation was meant to deliver is negated at precisely the point where output stops being information.","Mistral OCR4 and Unlimited-OCR are good models with significant technical advances behind them. The degeneration behavior described here does not define them; it identifies a specific failure condition that their current training has not addressed for this domain. The question is what a training pipeline designed to address it looks like.","In DharmaOCR, the answer is the DPO stage.","Supervised fine-tuning concentrates the model's resources on the target domain — the stage that builds the linguistic alignment described above. But SFT trains on individual token predictions: the model learns to produce the correct next token given the context preceding it. Under visual complexity, this creates the condition that produces degeneration. If an early token in the output diverges from the source document, each subsequent prediction is conditioned on that divergent state, and the output continues to drift. Repetition loops and incoherent sequences are an inherent characteristic in this context — they are the predictable result of an objective optimized step by step without accounting for the coherence of the full extraction.","DPO trains against a different signal. Where SFT trains token by token, DPO trains the model against the quality of complete outputs — teaching it to discriminate between competing responses based on the coherence of the full extraction rather than the accuracy of individual predictions. The effect is stabilizing: on documents where visual complexity would otherwise trigger drift, the model is less likely to commit to a divergent path, because its training penalized outputs that lost coherence at the extraction level.","The result is what the original benchmark demonstrated: lower degeneration rates alongside higher extraction accuracy, on the same documents where models without this training stage lose coherence.","The benchmark establishes what is true today. It is less precise about what will be true in two years.","It is possible — and likely — that newer models will eventually outperform the current DharmaOCR even in Brazilian Portuguese. Architectures will improve. Training techniques will advance. Datasets will expand. The field moves toward higher capability at every level, and there is no reason to expect that trajectory to stop. This is expected, and it is welcome.","What changes with each architectural generation is the ceiling on absolute performance. What does not change is the structural logic that determines which systems come closest to their ceiling in a given domain.","Available resources — compute, parameters, training data — are finite. They must be directed somewhere. A system that directs them at a single domain will extract more from them in that domain than a system distributing the same resources across many. This relationship holds regardless of how capable the generalist model has become. The gap between specialist and generalist may evolve as architectures improve. The structural dynamic does not reverse. We explored this principle in more depth in an earlier piece on why specialization continues to produce an advantage as AI systems advance (Why Specialization Is Inevitable：https://huggingface.co/blog/Dharma-AI/why-specialization-is-inevitable).","This is what shapes how Dharma approaches what comes next. The objective is not to defend the current model's benchmark position. It is to remain at the frontier of emerging techniques — new architectures, new training methods, new approaches to alignment and evaluation — and adapt them toward the same end: a specialized system for Brazilian Portuguese OCR that makes the best possible use of available resources, at the lowest possible cost and the shortest possible inference time.","Better tools do not work against specialization. They expand what it can achieve.","Three months ago, we showed that concentrating a model's training on a specific domain produces a measurable advantage over generalist systems, including ones that are newer and better-resourced. That advantage held. The same principle will determine how DharmaOCR continues to evolve — not by staying fixed, but by applying whatever progress the field makes to a domain that remains fixed.","Explore Dharma AI on Hugging Face ：https://huggingface.co/Dharma-AI to try our interactive demos ：https://huggingface.co/spaces/Dharma-AI/DharmaOCR-demo , download our open-source models ：https://huggingface.co/Dharma-AI/Dharma-OCR-LITE , and discover how specialized AI systems outperform general-purpose models in real enterprise applications.","Extract text and tables from document images"],"articleImages":[{"sourceUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/66a49c4207dd6d0ab485d3c2/zpZyMDJZMDPc-7cyqR292.jpeg","alt":"","afterParagraph":0,"url":"/media/articles/cmrni9h7w00c7bii5am7uma4i/53437c7ba15a4990.webp"}],"mediaStatus":"ok","articleBodyZh":["三个月前，我们发布了一篇关于DharmaOCR的论文：https://arxiv.org/abs/2604.14314，并开源了其中的一个模型：https://huggingface.co/Dharma-AI/Dharma-OCR-LITE。其目标非常明确：面向巴西葡萄牙语的光学字符识别。","训练流程分为两个阶段。第一阶段是监督式微调，利用来自不同来源、格式和复杂程度各异的葡萄牙语文件的大量集合。这一阶段将模型的权重与巴西葡萄牙语的特定词汇、句法和文档结构对齐——将表征能力集中在目标语言上，而不是分散在更广泛的多语言空间中。第二阶段应用了直接偏好优化（Direct Preference Optimization，DPO）：模型不仅仅在正确转录上进行训练，而是从相互竞争输出之间的比较偏好数据中学习，从而在推理时始终选择更优的提取结果。该阶段解决了不同的问题：不是准确性，而是稳定性。通过抑制会导致生成模型产生重复或不连贯输出的失败模式，DPO减少了推理时间和成本，并显著提升了模型在生产中提供结果的可靠性。","最终的综合结果是，在面向葡萄牙语的基准测试中，该模型以最低的退化率获得了最高的提取质量分数。两个阶段都是必要的。微调阶段建立了领域能力；DPO阶段确保了这种能力在模型容易失败的条件下依然稳健。","OCR模型发展迅速。但最初促使DharmaOCR设计的差距（在处理复杂文档的提取质量和生产环境下的模型稳定性方面）仍未消失。实际上，随着该领域的变化，这些差距反而变得更具启发性。","多模态生成模型的普及使基于语言模型的OCR广泛可用，随后出现的一系列微调OCR变体反映了这一采用速度之快。然而，这种普及并没有改变该技术的基本特性。所有建立在生成模型上的OCR系统都是概率性的。转录错误是这种概率技术固有的变量。模型之间的区别在于它们产生错误的数量和类型。这由两方面决定：模型的结构（其架构和参数数量）以及这些参数是如何为任务训练的。","架构和参数数量确定了模型能够学习的上限。训练决定了这种能力如何分配。","这一区别说明了专门化是一个结构性问题，而不是设计偏好。当模型在受限领域训练——单一语言、有限的文档类型或特定任务——其所有参数都专注于该特定任务。当模型被训练以涵盖更广泛的领域——例如处理N种语言的多语言模型——这些相同的参数必须分布在所有领域。分配不是线性的：神经元叠加原理意味着个别参数可以同时编码多个特征。但这种分配是真实的，其后果也是实实在在的。涵盖更多领域的模型在任何特定部分的投入都会减少。","DharmaOCR的训练是反向接受这一约束的。该模型并没有被设计为其他语言的最佳选择，也从未打算如此。作为交换，网络可用的每一个参数都可以面向巴西葡萄牙语的特定词汇、形态和正字法模式——对该领域的模型资源实现了最直接的使用。","这种集中化是相对于多语言及更广泛领域模型的固有优势的结构基础。这一优势并不依赖于拥有比竞争对手更大的架构或更复杂的训练程序——新架构和新训练技术提高了任何模型的能力。它依赖于这些资源被指向的方向：集中在一个领域，而不是分散在多个领域。","三个月后，更先进的模型已经出现。当这些模型更新且功能更强时，专门化的理由是否仍然成立，这是一个不同的问题。","DharmaOCR 论文发布三个月后，两个新的 OCR 模型在研究界引起了广泛关注：Mistral OCR4 和 Unlimited-OCR。它们都代表了真正的技术进步——新的训练技术、新的数据集，以及在多语言和各种基准评估中的出色表现。这类模型会提高 OCR 系统所预期达到的竞争标准。","当我们将两者在 DharmaOCR 基准上进行测试时——该评估是专门针对葡萄牙语设计的——结果是明确的。","DharmaOCR 得分 0.925。Mistral OCR4 得分 0.798。Unlimited-OCR 得分 0.7587。","差距很显著。Mistral OCR4 比 DharmaOCR 低约 13 分；Unlimited-OCR 比 DharmaOCR 低超过 16 分。两者均在我们的模型之后发布，且都拥有可观的研究资源。在 DharmaOCR 的基本设计决策是完全专注于葡萄牙语的任务中，专门化优势是可测量且显著的。","基准测试是核心发现。以下内容说明了为什么这种差距呈现出特定的形式。","处理非平凡的葡萄牙语文档会精确地揭示多语言模型容易出错的地方。ENEM 论文（巴西全国高中考试）结合了手写文本、特定于巴西葡萄牙语的词汇、专有名词和文化参考。它们正是语言特定训练能够产生收益的文档类型。","：https://cdn-uploads.huggingface.co/production/uploads/67b22efa9b36260fe286e99f/XVbylBZ_TyY98EvfkO9pT.png","图 1：用于基准评估的 ENEM 论文手稿及每个模型的输出。错误识别部分用红色标出。","Mistral OCR4 在评估这类文档时，将名字 Chico Buarque（巴西最著名的音乐家和诗人之一）识别为 “Chico Barque”。Unlimited-OCR 将同一个名字识别为 “chico bique”。当面对短语 “O Brasil não exclui, assimila”（“巴西不排斥，而是同化”，这是嵌入同一文档的 Chico Buarque 引用）时，Unlimited-OCR 输出了：“a dose de chico bique, 'o Brasil no exclu, eliminila.'”","这些并非随机错误。对巴西葡萄牙语接触不足的模型不会随意失败——它会在精确区分巴西葡萄牙语与更广泛多语种语料的词汇和专有名词上出现失误。Chico Buarque 并不是一个不为人知的参考；这个名字在全国范围内都是公认的。它在输出中系统性被损坏并不是边缘情况，而是一种诊断：显示模型训练存在的缺口。","DharmaOCR 在同一文档的评估中能够正确处理这些情况。原因直接：模型的训练集中在该语言空间，将资源集中在巴西葡萄牙语特有的词汇和专有名词分布上，而不是同时分散到多种语言。","这些例子说明了基准测试，而不是替代它。基准测试确定了差距的规模；这些例子显示差距为何集中在语言特定识别能力上，而非总体能力上。","然而，抽取准确性只是生产性能的一个维度。视觉困难下的稳定性是另一个维度——在实际操作中，失败的影响更为显著。","当生成模型遇到无法清晰解析的文档——小字体、扫描质量低、手写密集——它的输入信号就会出现不确定性。在主要以下一个标记预测目标训练的模型中，这里存在特定脆弱性：当视觉信号变得模糊时，模型可能会继续基于先前学习的模式生成内容，而非依赖源文档。结果是文本退化——输出重复、无序且与页面内容语义脱节。","面对小字体文档时，Mistral OCR4 生成的输出与文档所写内容没有关联。","：https://cdn-uploads.huggingface.co/production/uploads/69d815b52c6db28cfdfdd422/3YRHeWDsJ8twDlf7ArbZr.png","：https://cdn-uploads.huggingface.co/production/uploads/67b22efa9b36260fe286e99f/yBwDuIrqnY0xQ8RLY5Gbe.png","图 3：DharmaOCR 输出与 Mistral OCR4 退化输出","这不是源文本的低质量转录。这属于完全不同类别的失败。","其操作性后果不同于转录错误。错误的转录是可恢复的——它与源文档存在关系，原则上可以识别并纠正。退化的输出没有这种关系。它无法被纠正，因为没有可纠正的对象。对于依赖结构化 OCR 输出的后续流程——文档分类、信息提取、合规工作流——退化输出并不是“不准确的数据”。它是结构上无法使用的数据。自动化本应带来的效率在输出停止成为信息的那一刻被抵消。","Mistral OCR4 和 Unlimited-OCR 是性能良好的模型，背后有显著的技术进步。这里描述的退化行为并不定义它们；它仅指出当前训练在该领域未解决的特定失败条件。问题是，设计用于解决这个问题的训练流程应是什么样的。","在 DharmaOCR 中，答案是 DPO 阶段。","监督微调将模型的资源集中在目标领域——也就是构建上述语言对齐的阶段。但 SFT 是针对单个 token 进行训练：模型根据前面的上下文学习生成正确的下一个 token。在视觉复杂性下，这就产生了退化的条件。如果输出中的早期 token 与源文档偏离，则后续每个预测都基于该偏离状态，输出会继续漂移。在这种情况下，重复循环和不连贯序列是固有特性——它们是逐步优化目标而未考虑整体提取一致性的可预测结果。","DPO 针对不同的信号进行训练。SFT 是逐个 token 训练，而 DPO 则针对完整输出的质量训练模型——教它基于整个抽取的连贯性来区分竞争的响应，而不是单个预测的准确性。其效果具有稳定性：在视觉复杂性可能引发漂移的文档中，模型不太可能偏离路径，因为训练会惩罚在抽取层面失去连贯性的输出。","结果是原始基准测试所展示的：在同样的文档上，经过此训练阶段的模型表现出较低的退化率，同时抽取准确率更高，而没有此训练阶段的模型则会失去连贯性。","基准测试确立了今天的事实。但对两年后的情况则没有那么明确。","有可能——也很可能——更新的模型最终会在巴西葡萄牙语上超越当前的 DharmaOCR。架构会改进。训练技术会进步。数据集会扩展。整个领域在各个层面都朝着更高能力发展，没有理由认为这一趋势会停止。这是可以预期的，也是值得欢迎的。","每一代架构改变的是绝对性能的上限。不会改变的是决定在特定领域中哪些系统最接近其上限的结构逻辑。","可用资源——计算能力、参数量、训练数据——都是有限的，必须有所指向。将资源集中于单一领域的系统，在该领域的提取效率将高于将相同资源分散于多个领域的系统。无论通用模型能力如何增强，这种关系都成立。随着架构改进，专家模型与通用模型之间的差距可能会变化。结构性动态不会逆转。我们在之前一篇文章中更深入探讨了这一原理，讲述了为何随着 AI 系统的发展，专业化仍然会带来优势（Why Specialization Is Inevitable：https://huggingface.co/blog/Dharma-AI/why-specialization-is-inevitable）。","这就是塑造Dharma处理下一步的方法。目标不是去维护当前模型在基准测试中的位置，而是保持在新兴技术的前沿——新的架构、新的训练方法、对齐和评估的新方法——并将它们调整应用到同一个目标上：一个专用于巴西葡萄牙语OCR的系统，能够在最少成本和最短推理时间内最大限度地利用可用资源。","更好的工具不会与专业化对立。它们拓展了专业化可以实现的成就。","三个月前，我们展示了将模型的训练集中在特定领域上，相较于通用系统（包括更先进、资源更充足的系统）会产生可测量的优势。这种优势得到了保持。相同的原则将决定DharmaOCR如何继续发展——不是通过保持固定，而是通过将领域内的任何进展应用到这个固定的领域上。","在Hugging Face上探索Dharma AI：https://huggingface.co/Dharma-AI 试用我们的互动演示：https://huggingface.co/spaces/Dharma-AI/DharmaOCR-demo 下载我们的开源模型：https://huggingface.co/Dharma-AI/Dharma-OCR-LITE，了解专业化AI系统如何在实际企业应用中优于通用模型。","从文档图像中提取文本和表格"],"translationStatus":"translated","bodyOrigin":"source-page","editorial":{"summary":"Aioga 编辑摘要：DharmaOCR 在针对巴西葡萄牙语的基准测试中取得 0.925 分，而 Mistral OCR4 和 Unlimited-OCR 分别仅获 0.798 和 0.7587 分。 Aioga 将其归入「AI资讯」方向，重点关注它对真实使用和行业竞争的影响。","background":"背景分析：AI 行业动态需要结合来源、时间、实际可用性和后续反馈判断，标题或单次发布本身不能替代完整证据。","viewpoint":"Aioga 判断：这条动态更适合作为行业观察信号，当前信息足以建立线索，但不足以推导长期结论。","implications":"影响分析：对相关团队而言，短期应先核对来源、可用范围和实际成本，再判断是否值得接入或跟进。","nextStep":"后续观察：继续观察原文更新、官方说明、用户反馈和同类产品的后续动作。","evidenceRefs":["title","summary","articleBody"],"confidence":"medium","status":"published","aiGenerated":false,"autoApproved":true,"generatedBy":"rule-safe-fallback","generatedAt":"2026-07-23T09:13:51.041Z","sourceHash":"c0de58af7e376b02","validation":{"passed":true,"mode":"rule-safe-fallback","checks":["schema","length","source-attribution","no-html"]}},"tags":["AI资讯","Hugging Face：Blog（RSS）"],"translations":{"zh-CN":{"title":"Newer Models， Same Advantage","summary":"Newer Models， Same Advantage。该动态由 Hugging Face：Blog（RSS） 发布，Aioga 已同步发布时间、来源和原文入口，并将继续跟踪官方更新与行业反馈。","category":"AI资讯","source":"Hugging Face：Blog（RSS）","aggregationSource":"Hugging Face：Blog（RSS）","pageTitle":"Newer Models， Same Advantage - Aioga AI资讯","description":"Newer Models， Same Advantage。该动态由 Hugging Face：Blog（RSS） 发布，Aioga 已同步发布时间、来源和原文入口，并将继续跟踪官方更新与行业反馈。","url":"https://www.aioga.com/news/cmrni9h7w00c7bii5am7uma4i/"},"en":{"title":"Newer Models， Same Advantage","summary":"Aioga tracks this update from Hugging Face：Blog（RSS） under AI News. Newer Models， Same Advantage。该动态由 Hugging Face：Blog（RSS） 发布，Aioga 已同步发布时间、来源和原文入口，并将继续跟踪官方更新与行业反馈。","category":"AI News","source":"Hugging Face：Blog（RSS）","aggregationSource":"Hugging Face：Blog（RSS）","pageTitle":"Newer Models， Same Advantage - Aioga AI News","description":"Aioga tracks this update from Hugging Face：Blog（RSS） under AI News. Newer Models， Same Advantage。该动态由 Hugging Face：Blog（RSS） 发布，Aioga 已同步发布时间、来源和原文入口，并将继续跟踪官方更新与行业反馈。","url":"https://www.aioga.com/en/news/cmrni9h7w00c7bii5am7uma4i/"},"ja":{"title":"Newer Models， Same Advantage","summary":"Aiogaは「AIニュース」の動きとして、Hugging Face：Blog（RSS） からの更新を追跡しています。Newer Models， Same Advantage。该动态由 Hugging Face：Blog（RSS） 发布，Aioga 已同步发布时间、来源和原文入口，并将继续跟踪官方更新与行业反馈。","category":"AIニュース","source":"Hugging Face：Blog（RSS）","aggregationSource":"Hugging Face：Blog（RSS）","pageTitle":"Newer Models， Same Advantage - Aioga AIニュース","description":"Aiogaは「AIニュース」の動きとして、Hugging Face：Blog（RSS） からの更新を追跡しています。Newer Models， Same Advantage。该动态由 Hugging Face：Blog（RSS） 发布，Aioga 已同步发布时间、来源和原文入口，并将继续跟踪官方更新与行业反馈。","url":"https://www.aioga.com/ja/news/cmrni9h7w00c7bii5am7uma4i/"},"ko":{"title":"Newer Models， Same Advantage","summary":"Aioga는 Hugging Face：Blog（RSS）의 업데이트를 AI 뉴스 흐름으로 추적합니다. Newer Models， Same Advantage。该动态由 Hugging Face：Blog（RSS） 发布，Aioga 已同步发布时间、来源和原文入口，并将继续跟踪官方更新与行业反馈。","category":"AI 뉴스","source":"Hugging Face：Blog（RSS）","aggregationSource":"Hugging Face：Blog（RSS）","pageTitle":"Newer Models， Same Advantage - Aioga AI 뉴스","description":"Aioga는 Hugging Face：Blog（RSS）의 업데이트를 AI 뉴스 흐름으로 추적합니다. Newer Models， Same Advantage。该动态由 Hugging Face：Blog（RSS） 发布，Aioga 已同步发布时间、来源和原文入口，并将继续跟踪官方更新与行业反馈。","url":"https://www.aioga.com/ko/news/cmrni9h7w00c7bii5am7uma4i/"},"es":{"title":"Newer Models， Same Advantage","summary":"Aioga sigue esta actualización de Hugging Face：Blog（RSS） dentro de Noticias IA. Newer Models， Same Advantage。该动态由 Hugging Face：Blog（RSS） 发布，Aioga 已同步发布时间、来源和原文入口，并将继续跟踪官方更新与行业反馈。","category":"Noticias IA","source":"Hugging Face：Blog（RSS）","aggregationSource":"Hugging Face：Blog（RSS）","pageTitle":"Newer Models， Same Advantage - Aioga Noticias de IA","description":"Aioga sigue esta actualización de Hugging Face：Blog（RSS） dentro de Noticias IA. Newer Models， Same Advantage。该动态由 Hugging Face：Blog（RSS） 发布，Aioga 已同步发布时间、来源和原文入口，并将继续跟踪官方更新与行业反馈。","url":"https://www.aioga.com/es/news/cmrni9h7w00c7bii5am7uma4i/"},"fr":{"title":"Newer Models， Same Advantage","summary":"Aioga suit cette mise à jour de Hugging Face：Blog（RSS） dans la catégorie Actu IA. Newer Models， Same Advantage。该动态由 Hugging Face：Blog（RSS） 发布，Aioga 已同步发布时间、来源和原文入口，并将继续跟踪官方更新与行业反馈。","category":"Actu IA","source":"Hugging Face：Blog（RSS）","aggregationSource":"Hugging Face：Blog（RSS）","pageTitle":"Newer Models， Same Advantage - Aioga Actualités IA","description":"Aioga suit cette mise à jour de Hugging Face：Blog（RSS） dans la catégorie Actu IA. Newer Models， Same Advantage。该动态由 Hugging Face：Blog（RSS） 发布，Aioga 已同步发布时间、来源和原文入口，并将继续跟踪官方更新与行业反馈。","url":"https://www.aioga.com/fr/news/cmrni9h7w00c7bii5am7uma4i/"},"de":{"title":"Newer Models， Same Advantage","summary":"Aioga tracks this update from Hugging Face：Blog（RSS） under AI资讯. Newer Models， Same Advantage。该来源目前仅提供简短信息，Aioga 已保留发布时间、来源和原文入口，并将继续跟踪后续更新。来源：Hugging Face：Blog（RSS）。","category":"AI资讯","source":"Hugging Face：Blog（RSS）","aggregationSource":"Hugging Face：Blog（RSS）","pageTitle":"Newer Models， Same Advantage - Aioga KI-News","description":"Aioga tracks this update from Hugging Face：Blog（RSS） under AI资讯. Newer Models， Same Advantage。该来源目前仅提供简短信息，Aioga 已保留发布时间、来源和原文入口，并将继续跟踪后续更新。来源：Hugging Face：Blog（RSS）。","url":"https://www.aioga.com/de/news/cmrni9h7w00c7bii5am7uma4i/"},"pt-BR":{"title":"Newer Models， Same Advantage","summary":"Aioga tracks this update from Hugging Face：Blog（RSS） under AI资讯. Newer Models， Same Advantage。该来源目前仅提供简短信息，Aioga 已保留发布时间、来源和原文入口，并将继续跟踪后续更新。来源：Hugging Face：Blog（RSS）。","category":"AI资讯","source":"Hugging Face：Blog（RSS）","aggregationSource":"Hugging Face：Blog（RSS）","pageTitle":"Newer Models， Same Advantage - Aioga Notícias de IA","description":"Aioga tracks this update from Hugging Face：Blog（RSS） under AI资讯. Newer Models， Same Advantage。该来源目前仅提供简短信息，Aioga 已保留发布时间、来源和原文入口，并将继续跟踪后续更新。来源：Hugging Face：Blog（RSS）。","url":"https://www.aioga.com/pt-BR/news/cmrni9h7w00c7bii5am7uma4i/"},"ru":{"title":"Newer Models， Same Advantage","summary":"Aioga tracks this update from Hugging Face：Blog（RSS） under AI资讯. Newer Models， Same Advantage。该来源目前仅提供简短信息，Aioga 已保留发布时间、来源和原文入口，并将继续跟踪后续更新。来源：Hugging Face：Blog（RSS）。","category":"AI资讯","source":"Hugging Face：Blog（RSS）","aggregationSource":"Hugging Face：Blog（RSS）","pageTitle":"Newer Models， Same Advantage - Aioga Новости ИИ","description":"Aioga tracks this update from Hugging Face：Blog（RSS） under AI资讯. Newer Models， Same Advantage。该来源目前仅提供简短信息，Aioga 已保留发布时间、来源和原文入口，并将继续跟踪后续更新。来源：Hugging Face：Blog（RSS）。","url":"https://www.aioga.com/ru/news/cmrni9h7w00c7bii5am7uma4i/"},"ar":{"title":"Newer Models， Same Advantage","summary":"Aioga tracks this update from Hugging Face：Blog（RSS） under AI资讯. Newer Models， Same Advantage。该来源目前仅提供简短信息，Aioga 已保留发布时间、来源和原文入口，并将继续跟踪后续更新。来源：Hugging Face：Blog（RSS）。","category":"AI资讯","source":"Hugging Face：Blog（RSS）","aggregationSource":"Hugging Face：Blog（RSS）","pageTitle":"Newer Models， Same Advantage - Aioga أخبار الذكاء الاصطناعي","description":"Aioga tracks this update from Hugging Face：Blog（RSS） under AI资讯. Newer Models， Same Advantage。该来源目前仅提供简短信息，Aioga 已保留发布时间、来源和原文入口，并将继续跟踪后续更新。来源：Hugging Face：Blog（RSS）。","url":"https://www.aioga.com/ar/news/cmrni9h7w00c7bii5am7uma4i/"},"hi":{"title":"Newer Models， Same Advantage","summary":"Aioga tracks this update from Hugging Face：Blog（RSS） under AI资讯. Newer Models， Same Advantage。该来源目前仅提供简短信息，Aioga 已保留发布时间、来源和原文入口，并将继续跟踪后续更新。来源：Hugging Face：Blog（RSS）。","category":"AI资讯","source":"Hugging Face：Blog（RSS）","aggregationSource":"Hugging Face：Blog（RSS）","pageTitle":"Newer Models， Same Advantage - Aioga AI समाचार","description":"Aioga tracks this update from Hugging Face：Blog（RSS） under AI资讯. Newer Models， Same Advantage。该来源目前仅提供简短信息，Aioga 已保留发布时间、来源和原文入口，并将继续跟踪后续更新。来源：Hugging Face：Blog（RSS）。","url":"https://www.aioga.com/hi/news/cmrni9h7w00c7bii5am7uma4i/"},"it":{"title":"Newer Models， Same Advantage","summary":"Aioga tracks this update from Hugging Face：Blog（RSS） under AI资讯. Newer Models， Same Advantage。该来源目前仅提供简短信息，Aioga 已保留发布时间、来源和原文入口，并将继续跟踪后续更新。来源：Hugging Face：Blog（RSS）。","category":"AI资讯","source":"Hugging Face：Blog（RSS）","aggregationSource":"Hugging Face：Blog（RSS）","pageTitle":"Newer Models， Same Advantage - Aioga Notizie IA","description":"Aioga tracks this update from Hugging Face：Blog（RSS） under AI资讯. Newer Models， Same Advantage。该来源目前仅提供简短信息，Aioga 已保留发布时间、来源和原文入口，并将继续跟踪后续更新。来源：Hugging Face：Blog（RSS）。","url":"https://www.aioga.com/it/news/cmrni9h7w00c7bii5am7uma4i/"},"nl":{"title":"Newer Models， Same Advantage","summary":"Aioga tracks this update from Hugging Face：Blog（RSS） under AI资讯. Newer Models， Same Advantage。该来源目前仅提供简短信息，Aioga 已保留发布时间、来源和原文入口，并将继续跟踪后续更新。来源：Hugging Face：Blog（RSS）。","category":"AI资讯","source":"Hugging Face：Blog（RSS）","aggregationSource":"Hugging Face：Blog（RSS）","pageTitle":"Newer Models， Same Advantage - Aioga AI-nieuws","description":"Aioga tracks this update from Hugging Face：Blog（RSS） under AI资讯. Newer Models， Same Advantage。该来源目前仅提供简短信息，Aioga 已保留发布时间、来源和原文入口，并将继续跟踪后续更新。来源：Hugging Face：Blog（RSS）。","url":"https://www.aioga.com/nl/news/cmrni9h7w00c7bii5am7uma4i/"},"tr":{"title":"Newer Models， Same Advantage","summary":"Aioga tracks this update from Hugging Face：Blog（RSS） under AI资讯. Newer Models， Same Advantage。该来源目前仅提供简短信息，Aioga 已保留发布时间、来源和原文入口，并将继续跟踪后续更新。来源：Hugging Face：Blog（RSS）。","category":"AI资讯","source":"Hugging Face：Blog（RSS）","aggregationSource":"Hugging Face：Blog（RSS）","pageTitle":"Newer Models， Same Advantage - Aioga AI Haberleri","description":"Aioga tracks this update from Hugging Face：Blog（RSS） under AI资讯. Newer Models， Same Advantage。该来源目前仅提供简短信息，Aioga 已保留发布时间、来源和原文入口，并将继续跟踪后续更新。来源：Hugging Face：Blog（RSS）。","url":"https://www.aioga.com/tr/news/cmrni9h7w00c7bii5am7uma4i/"},"vi":{"title":"Newer Models， Same Advantage","summary":"Aioga tracks this update from Hugging Face：Blog（RSS） under AI资讯. Newer Models， Same Advantage。该来源目前仅提供简短信息，Aioga 已保留发布时间、来源和原文入口，并将继续跟踪后续更新。来源：Hugging Face：Blog（RSS）。","category":"AI资讯","source":"Hugging Face：Blog（RSS）","aggregationSource":"Hugging Face：Blog（RSS）","pageTitle":"Newer Models， Same Advantage - Tin tức AI Aioga","description":"Aioga tracks this update from Hugging Face：Blog（RSS） under AI资讯. Newer Models， Same Advantage。该来源目前仅提供简短信息，Aioga 已保留发布时间、来源和原文入口，并将继续跟踪后续更新。来源：Hugging Face：Blog（RSS）。","url":"https://www.aioga.com/vi/news/cmrni9h7w00c7bii5am7uma4i/"},"id":{"title":"Newer Models， Same Advantage","summary":"Aioga tracks this update from Hugging Face：Blog（RSS） under AI资讯. Newer Models， Same Advantage。该来源目前仅提供简短信息，Aioga 已保留发布时间、来源和原文入口，并将继续跟踪后续更新。来源：Hugging Face：Blog（RSS）。","category":"AI资讯","source":"Hugging Face：Blog（RSS）","aggregationSource":"Hugging Face：Blog（RSS）","pageTitle":"Newer Models， Same Advantage - Berita AI Aioga","description":"Aioga tracks this update from Hugging Face：Blog（RSS） under AI资讯. Newer Models， Same Advantage。该来源目前仅提供简短信息，Aioga 已保留发布时间、来源和原文入口，并将继续跟踪后续更新。来源：Hugging Face：Blog（RSS）。","url":"https://www.aioga.com/id/news/cmrni9h7w00c7bii5am7uma4i/"},"th":{"title":"Newer Models， Same Advantage","summary":"Aioga tracks this update from Hugging Face：Blog（RSS） under AI资讯. Newer Models， Same Advantage。该来源目前仅提供简短信息，Aioga 已保留发布时间、来源和原文入口，并将继续跟踪后续更新。来源：Hugging Face：Blog（RSS）。","category":"AI资讯","source":"Hugging Face：Blog（RSS）","aggregationSource":"Hugging Face：Blog（RSS）","pageTitle":"Newer Models， Same Advantage - ข่าว AI Aioga","description":"Aioga tracks this update from Hugging Face：Blog（RSS） under AI资讯. Newer Models， Same Advantage。该来源目前仅提供简短信息，Aioga 已保留发布时间、来源和原文入口，并将继续跟踪后续更新。来源：Hugging Face：Blog（RSS）。","url":"https://www.aioga.com/th/news/cmrni9h7w00c7bii5am7uma4i/"},"pl":{"title":"Newer Models， Same Advantage","summary":"Aioga tracks this update from Hugging Face：Blog（RSS） under AI资讯. Newer Models， Same Advantage。该来源目前仅提供简短信息，Aioga 已保留发布时间、来源和原文入口，并将继续跟踪后续更新。来源：Hugging Face：Blog（RSS）。","category":"AI资讯","source":"Hugging Face：Blog（RSS）","aggregationSource":"Hugging Face：Blog（RSS）","pageTitle":"Newer Models， Same Advantage - Aioga Wiadomości AI","description":"Aioga tracks this update from Hugging Face：Blog（RSS） under AI资讯. Newer Models， Same Advantage。该来源目前仅提供简短信息，Aioga 已保留发布时间、来源和原文入口，并将继续跟踪后续更新。来源：Hugging Face：Blog（RSS）。","url":"https://www.aioga.com/pl/news/cmrni9h7w00c7bii5am7uma4i/"}}}}