{"ID":2826164,"CreatedAt":"2026-06-01T04:54:23.091178241Z","UpdatedAt":"2026-06-01T04:54:23.091178241Z","DeletedAt":null,"paper_url":"https://arxiv.org/abs/2512.19171","arxiv_id":"2512.19171","title":"JEPA-Reasoner: Decoupling Latent Reasoning from Token Generation","abstract":"Current autoregressive language models couple high-level reasoning and low-level token generation into a single sequential process, making the reasoning trajectory vulnerable to compounding expression errors. We propose JEPA-Reasoner, a novel architectural paradigm that decouples these tasks using a Joint-Embedding Predictive Architecture (JEPA) for pure latent-space reasoning and a separate Talker module for linguistic reconstruction. By isolating the reasoning engine from the discrete token-sampling process, our architecture enables: (1) Error Containment, where token-level failures cannot propagate into the latent reasoning chain; (2) Continuous Guidance, providing the generator with access to the entire lossless reasoning trajectory; and (3) Representation of Uncertainty, allowing the model to maintain multiple hypotheses via mixed latent vectors. Controlled experiments on synthetic and natural language tasks demonstrate that this decoupling enables a 0.9B model to achieve a 149.5\\% improvement in 8-shot GSM8K accuracy over a coupled Transformer baseline trained on identical data. This work shifts the focus from scaling coupled models to investigating decoupled architectures as a more robust foundation for complex reasoning.","short_abstract":"Current autoregressive language models couple high-level reasoning and low-level token generation into a single sequential process, making the reasoning trajectory vulnerable to compounding expression errors. We propose JEPA-Reasoner, a novel architectural paradigm that decouples these tasks using a Joint-Embedding Pre...","url_abs":"https://arxiv.org/abs/2512.19171","url_pdf":"https://arxiv.org/pdf/2512.19171v3","authors":"[\"Bingyang Kelvin Liu\",\"Ziyu Patrick Chen\",\"David P. Woodruff\"]","published":"2025-12-22T09:05:06Z","proceeding":"cs.CL","tasks":"[\"cs.CL\"]","methods":"[\"Transformer\",\"Language Model\"]","has_code":false}