HTML 77.2%
TypeScript 10.5%
Python 9.6%
JavaScript 2.5%
12 <rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom">3 <channel>4 <title>Apple Machine Learning Research</title>5 <link>https://machinelearning.apple.com</link>6 <description>Apple machine learning teams are engaged in state of the art research in machine learning and artificial intelligence. Learn about the latest advancements.</description>7 <language>en</language>8 <lastBuildDate>Fri, 11 Sep 2026 00:00:00 GMT</lastBuildDate>9 <atom:link href="https://machinelearning.apple.com/rss.xml" rel="self" type="application/rss+xml"/>10 11 <item>12 <guid>discosign-gloss-translation</guid>13 <title>DiscoSign: Discourse-Aware Text to Sign Language Gloss Translation</title>14 <link>https://machinelearning.apple.com/research/discosign-gloss-translation</link>15 <description>Sign language processing systems have traditionally operated at the sentence level, ignoring critical discourse phenomena fundamental to sign language comprehension. We introduce DiscoSign, a computational approach for discourse-aware text to sign language gloss translation grounded in linguistic research. We address three key phenomena within our modular Large Language Model (LLM)-based translation framework: (i) spatial coreference resolution, where entities maintain consistent spatial locations throughout discourse; (ii) Question-Answer Clauses (QACs), pseudocleft structures serving…</description>16 <pubDate>Fri, 11 Sep 2026 00:00:00 GMT</pubDate>17 </item>1819 <item>20 <guid>video-caption-quality</guid>21 <title>Putting Captions to the Test: Evaluating Video Caption Quality through Multiple-Choice Question Answering</title>22 <link>https://machinelearning.apple.com/research/video-caption-quality</link>23 <description>Evaluating video captioning remains a critical challenge for Visual Large Language Models (VLLMs). Existing metrics primarily rely on matching generated text against ground-truth references. This paradigm suffers from the “one-to-many” nature of video description, where high-quality captions are often penalized for lexical mismatches or valid shifts in visual focus. Furthermore, such assessments are typically one-dimensional, failing to provide a fine-grained analysis of caption quality. To address this, we redefine caption quality via information fidelity: A caption must maximize the coverage…</description>24 <pubDate>Fri, 11 Sep 2026 00:00:00 GMT</pubDate>25 </item>2627 <item>28 <guid>simpledesign-protein-codesign</guid>29 <title>SimpleDesign: A Joint Model for Protein Sequence and Structure Codesign</title>30 <link>https://machinelearning.apple.com/research/simpledesign-protein-codesign</link>31 <description>Proteins are fundamental to biological processes, with their function determined by the complex interplay between the amino acid sequence and the three-dimensional structure. Developing generative models capable of understanding this intrinsically multi-modal relationship is crucial for fields like drug discovery and protein engineering. Existing models often rely on a multi-stage training process where autoencoders that tokenize data into latent representations are trained in a first stage. Secondly, a generative model is trained on the latent representation of the autoencoder(s), i.e…</description>32 <pubDate>Fri, 11 Sep 2026 00:00:00 GMT</pubDate>33 </item>3435 <item>36 <guid>refactor-vla-motor-programs</guid>37 <title>REFACTOR-VLA: Unsupervised Library Learning of Typed Motor Programs</title>38 <link>https://machinelearning.apple.com/research/refactor-vla-motor-programs</link>39 <description>Most current vision-language-action (VLA) models—such as OpenVLA, π0, RT-2, and RDT-1B—are “monolithic.” This means they generate raw motor commands or very short sequences of actions, without organizing behaviors into reusable, well-defined abstractions. As a result, these models perform poorly on long-horizon (multi-step) tasks, and it’s difficult to interpret what they have learned. Existing approaches for discovering skills often avoid the core problem of deciding when two action sequences are “behaviorally equivalent.” For example, AtomicVLA and AtomSkill group action sequences by…</description>40 <pubDate>Wed, 02 Sep 2026 00:00:00 GMT</pubDate>41 </item>4243 <item>44 <guid>llms-not-consistently-bayesian</guid>45 <title>LLMs Are Not (Consistently) Bayesian: Quantifying Internal (In)consistencies of LLMs’ Probabilistic Beliefs</title>46 <link>https://machinelearning.apple.com/research/llms-not-consistently-bayesian</link>47 <description>Modern AI systems are being deployed in complex domains such as medicine, science, and law, where there is often not a single correct answer given the observed evidence. Such systems must be able to represent and update uncertain beliefs about the world as new evidence arrives to make rational decisions. We introduce the novel technique of studying LLMs as information processing rules and utilize the information processing gap—the deviation from Bayes updates—to study the internal (in)consistencies of how LLMs update their probabilistic beliefs from evidence. Our extensive experiments evaluate…</description>48 <pubDate>Fri, 28 Aug 2026 00:00:00 GMT</pubDate>49 </item>5051 <item>52 <guid>agent-seer-synthesizing-scenarios</guid>53 <title>Agent Seer: Synthesizing Scenarios from Specification Understanding</title>54 <link>https://machinelearning.apple.com/research/agent-seer-synthesizing-scenarios</link>55 <description>Evaluating AI agents that use external tools requires realistic test scenarios that capture how practitioners compose tools and iterate across conversation turns. Constructing such scenarios by hand demands deep domain expertise, does not scale across tool ecosystems, and produces static benchmarks that cannot track evolving APIs. We observe that tool specifications—function names, natural-language descriptions, and typed parameter schemas—already encode sufficient semantic information to synthesize realistic evaluation scenarios without manual curation or live tool execution. Agent Seer…</description>56 <pubDate>Fri, 28 Aug 2026 00:00:00 GMT</pubDate>57 </item>5859 <item>60 <guid>rubric-based-alignment</guid>61 <title>From Preferences to Principles: Rubric-Based Alignment for Grounded Knowledge Answers</title>62 <link>https://machinelearning.apple.com/research/rubric-based-alignment</link>63 <description>Designing effective reward signals for open-domain question answering is challenging because high-quality responses must simultaneously satisfy multiple aspects of answer quality that are difficult to capture with a holistic scalar objective. We introduce a rubric-based reward framework that generates query-specific rubrics grounded in retrieved evidence and decomposed into multiple quality dimensions, providing fine-grained supervision during post-training. Averaged across three evaluation axes (composition, grounding, and instruction-following), our approach improves over the…</description>64 <pubDate>Thu, 27 Aug 2026 00:00:00 GMT</pubDate>65 </item>6667 <item>68 <guid>proof-gen-optimized-distillation</guid>69 <title>PROOF-Gen: From Optimized Data to Better Distillation</title>70 <link>https://machinelearning.apple.com/research/proof-gen-optimized-distillation</link>71 <description>Supervised fine-tuning on teacher-generated trajectories is the standard first stage for distilling tool-calling capabilities into deployable models. Post-training pipelines that drive shipped tool-calling agents re-run this stage on a daily or weekly cadence, paying the frontier-teacher cost each cycle, yet the mechanism is generate-and-filter (keep the teacher’s passing trajectories, discard the rest) and each cycle leaves behind the same hard scenarios because failures supply no signal. On τ 2-bench, 57% of teacher trials fail, two-thirds of them near-misses (most tool calls correct, undone…</description>72 <pubDate>Wed, 26 Aug 2026 00:00:00 GMT</pubDate>73 </item>7475 <item>76 <guid>idea-prune-pipeline</guid>77 <title>IDEA Prune: An Integrated Enlarge-and-Prune Pipeline in Generative Language Model Pretraining</title>78 <link>https://machinelearning.apple.com/research/idea-prune-pipeline</link>79 <description>Recent advancements in large language models have intensified the need for efficient and deployable models within limited inference budgets. Structured pruning pipelines have shown promise in token efficiency compared to training target-size models from scratch. In this paper, we advocate incorporating enlarged model pretraining, which is often ignored in previous works, into pruning. We study the enlarge-and-prune pipeline as an integrated system to address two critical questions: whether it is worth pretraining an enlarged model even when the model is never deployed, and how to optimize the…</description>80 <pubDate>Wed, 26 Aug 2026 00:00:00 GMT</pubDate>81 </item>8283 <item>84 <guid>relightable-gaussians-3d-generation</guid>85 <title>Luce: Relightable Gaussians for 3D Asset Generation</title>86 <link>https://machinelearning.apple.com/research/relightable-gaussians-3d-generation</link>87 <description>High-fidelity image-to-3D generation requires a 3D representation that captures both geometry and appearance. To support relighting and integration into standard rendering pipelines, the representation should include physically based rendering (PBR) modalities such as albedo, metallic-roughness, and surface normals. We propose Luce, a 3D representation that unifies geometry and PBR materials within a voxelized multimodal Gaussian cloud, using dedicated Gaussian primitives for each modality. A variational autoencoder compresses this representation into a unified material-aware latent space. A…</description>88 <pubDate>Wed, 26 Aug 2026 00:00:00 GMT</pubDate>89 </item>9091 </channel>92 </rss>93