{"slug": "bringing-laya-onto-an-iphone-a-local-decision-model-experiment", "title": "Bringing Laya onto an iPhone: a local decision-model experiment", "summary": "A developer built a Simplified Chinese iPhone app, 他爱我吗, that runs a Core ML conversion of the laya-multilingual model entirely on-device to score candidate answers to relationship questions such as affection, no affection, and uncertainty. The app uses local SenseVoice transcription and template-based responses rather than a hosted chatbot, and keeps history on the device. The developer stresses that displayed scores are relative preferences among supplied options, not probabilities that a person is in love, and proposes reproducible evaluation directions instead of an accuracy claim.", "body_md": "Disclosure: I am the maker of 他爱我吗. This article was prepared with AI assistance from the project's implementation notes.\n\nA question such as ‘Does he like me?’ looks simple in an interface. It is much harder to decide what an answer should mean. A model cannot see the other person's thoughts, and an attractive percentage does not change that.\n\nI built a Chinese-language iPhone app to explore a narrower interaction: describe a situation, compare candidate answers, and leave room for uncertainty. Here is the architecture and the product boundary behind that experiment.\n\nThe app uses a Core ML conversion of laya-multilingual to compare supplied answer options for the question and context. For a relationship question, those options might be affection, no affection, and uncertainty. The wording needs to fit the actual question.\n\nThis is not an unrestricted chatbot generating long responses through a hosted language model. Short conclusions and suggestions use templates. Chinese speech transcription uses SenseVoice locally, and history remains on the device.\n\n```\nChinese text or speech\n        ↓\nLocal transcription, when needed\n        ↓\nQuestion + context + candidate answers\n        ↓\nLaya through Core ML\n        ↓\nRelative answer scores + template-based response\n```\n\nKeeping analysis on-device is useful when the input is personal. It also shifts responsibilities to the app: model packaging, storage, loading, and keeping the interface responsive become part of the mobile experience. I am not claiming a latency or battery benchmark here; those require measured device tests.\n\nA displayed answer score is a relative preference among the options given to the model. It is not evidence that a person has a corresponding probability of being in love.\n\nAn uncertain result belongs in the interaction rather than being treated as an error. More context can help frame a question, but it still cannot establish another person's internal state. The interface should make that limit visible close to the result.\n\nThe image below is an actual Chinese-language result screen. It illustrates the interface, not a validated prediction about anyone.\n\nThere are useful questions to investigate before making stronger claims about this type of product:\n\nThese are proposed evaluation directions, not completed experimental results. A reproducible set of cases would be more useful than an unsupported accuracy number.\n\n[Laya's upstream project](https://github.com/NandhaKishorM/laya) describes a non-autoregressive System 1 decision engine and an optional Jev-compatible server interface at `/v1/systemone`. That is related technical context, not the architecture of this iOS app: the app uses local Laya inference and does not call the Jev API or run that server.\n\nThe conversion project is available at [laya-coreml](https://github.com/mizorewww/laya-coreml). The app itself is not being presented as an open-source project.\n\nThe current interface is Simplified Chinese. The [project website](https://doesheloveme.tennote.site/en/) shows the screens and limitations. If you are working on a local decision model, I would be interested in comparing evaluation approaches—especially how you prevent users from interpreting a relative score as certainty.", "url": "https://wpnews.pro/news/bringing-laya-onto-an-iphone-a-local-decision-model-experiment", "canonical_source": "https://dev.to/yang9527/bringing-laya-onto-an-iphone-a-local-decision-model-experiment-4af0", "published_at": "2026-09-27 11:28:34+00:00", "updated_at": "2026-09-27 12:01:00.277018+00:00", "lang": "en", "topics": ["ai-products", "ai-tools", "natural-language-processing", "machine-learning"], "entities": ["他爱我吗", "laya-multilingual", "Core ML", "SenseVoice", "laya-coreml", "Jev", "Apple"], "also_reported_by": [], "alternates": {"html": "https://wpnews.pro/news/bringing-laya-onto-an-iphone-a-local-decision-model-experiment", "markdown": "https://wpnews.pro/news/bringing-laya-onto-an-iphone-a-local-decision-model-experiment.md", "text": "https://wpnews.pro/news/bringing-laya-onto-an-iphone-a-local-decision-model-experiment.txt", "jsonld": "https://wpnews.pro/news/bringing-laya-onto-an-iphone-a-local-decision-model-experiment.jsonld"}}