{"name":"lingbot-touch","repo":"Robbyant/lingbot-touch","url":"https://midorreal.com/project/robbyant-lingbot-touch","repository":"https://github.com/Robbyant/lingbot-touch","description":"Learning Heterogeneous Tactile Representation for General Dexterous Manipulation","facts":{"license":{"value":"Apache-2.0","source":"https://github.com/Robbyant/lingbot-touch","checked":"22 Nov 2026"},"language":{"value":"Python","source":"https://github.com/Robbyant/lingbot-touch","checked":"22 Nov 2026"},"releases":{"value":"0","source":"https://github.com/Robbyant/lingbot-touch/releases","checked":"22 Nov 2026"},"contributors":{"value":"3","source":"https://github.com/Robbyant/lingbot-touch/graphs/contributors","checked":"22 Nov 2026"},"archived":{"value":"no","source":"https://github.com/Robbyant/lingbot-touch","checked":"22 Nov 2026"}},"stars":797.0,"writeup":{"what_it_is":"LingBot-Touch is a JEPA-based tactile representation model. It uses a shared Transformer and latent prediction to learn from heterogeneous tactile sensor signals, and is intended for dexterous robotic manipulation. The repository provides a frozen target encoder and utilities for extracting token-level tactile features.","audience":"Researchers and engineers working on tactile sensing and dexterous robotic manipulation.","claims":[{"kind":"broad","claim":"The model supports high-frequency downstream inference with a lightweight architecture.","excerpt":"Even with its smallest ViT backbone, LingBot-Touch supports high-frequency downstream inference with a lightweight architecture.","status":"too_broad"},{"kind":"specific","claim":"Pretraining supports specific point counts per finger for magnetic, piezoresistive and visuotactile sensing.","excerpt":"Pretraining explicitly supports **96, 76, 64, 48, 32, 24** points per finger for magnetic sensing","status":"not_checked"},{"kind":"specific","claim":"The model transfers to downstream tactile sensors and real-world robotic hands without additional training.","excerpt":"LingBot-Touch transfers to downstream tactile sensors and real-world dexterous robotic hands without any additional training.","status":"not_checked"},{"kind":"broad","claim":"Performance improves as the amount of tactile training data grows.","excerpt":"Scaling the amount of tactile training data yields continued performance improvements","status":"too_broad"},{"kind":"specific","claim":"Heterogeneous tactile data from multiple sources is used within a unified latent space.","excerpt":"Heterogeneous tactile data from multiple sources mutually enhance representation learning within a unified latent space.","status":"not_checked"}],"alternatives":[],"written_by":"AI, from the project's README","date":"2026-11-13"},"why_now":null,"owner_supplied":null,"score":{"verdict":"too_early","hype":71,"reality":24,"gap":47,"momentum":null,"confidence":60,"version":"score-2.0","rule":"verdict-2.0","calculated_at":"2026-11-22T07:00:00+00:00","verdict_source":"rule"},"early_read":{"estimate":true,"lean":"too_close","gap_low":10,"gap_high":52,"rule":"early-2.0"},"changes":[{"date":"2026-11-09T09:02:29+00:00","text":"Repository created","source":"https://github.com/Robbyant/lingbot-touch"}],"cite":"lingbot-touch is a GitHub project at https://github.com/Robbyant/lingbot-touch written mainly in Python, described there as \"Learning Heterogeneous Tactile Representation for General Dexterous Manipulation\".","attribution":"Data from Mid or Real, https://midorreal.com/project/robbyant-lingbot-touch, checked 22 November 2026. Attribution with a link is required.","still_real":null}