Matthew Jörke

22 posts

Matthew Jörke

Matthew Jörke

@mjoerke

PhD Student @stanford CS | HCI/AI, health behavior change | he/him

Stanford, CA Katılım Mayıs 2024
87 Takip Edilen119 Takipçiler
Matthew Jörke retweetledi
Michael Y. Li
Michael Y. Li@michaelyli__·
Can a language model learn, end-to-end, what to keep in its own KV cache and what to throw away? Can it learn to forget while it learns to reason? Deep learning's central lesson: capability emerges from end-to-end optimization, not heuristics/strong inductive biases. But for efficiency, we rely heavily on hand-designed approaches. 🗑️ Introducing Neural Garbage Collection (NGC): we train a language model to jointly reason and manage its own KV cache, using reinforcement learning with outcome-based task reward alone. No SFT, no proxy objectives, no summarization in natural language. New paper with @jubayer_hamid, Emily Fox, and @noahdgoodman!
Michael Y. Li tweet media
English
25
135
901
158.4K
Matthew Jörke
Matthew Jörke@mjoerke·
@FerryLee_AIPOCH The app integrates with Apple's HealthKit API so while we used Apple Watches in our study, the platform itself is not tied to any particular wearable! If your wearable/smart device can read/write to HealthKit, Bloom can read that data too
English
0
0
0
26
FerryLee
FerryLee@FerryLee_AIPOCH·
@mjoerke Love the UI addition, it’s a great solution. At least now we don’t have to keep telling AI “check the context, I already told you that” 😅 Any plans to connect with other devices too, like smartwatches or smart scales?
English
1
0
0
33
Matthew Jörke retweetledi
Aishwarya Mandyam @ ✈️ ICLR 2026
✨I'm on the research scientist and postdoc job market! I'll be graduating from my PhD this academic year with a thesis that focuses on reinforcement learning and healthcare. ✨
English
12
45
337
52K
Matthew Jörke
Matthew Jörke@mjoerke·
In a counterfactual comparison to vanilla GPT4, GPTCoach is more consistent with motivational interviewing, asking more open-ended questions and giving advice with permission.
Matthew Jörke tweet media
English
1
1
5
619
Matthew Jörke
Matthew Jörke@mjoerke·
In a user study with 16 participants, we find that GPTCoach can adhere to motivational interviewing principles and contextualize a user's wearable data to their unique circumstances. Participants also appreciated its supportive and non-judgmental tone.
English
1
1
6
647
Matthew Jörke
Matthew Jörke@mjoerke·
We built GPTCoach, a GPT4-based chatbot that implements an evidence-based health coaching program, uses counseling strategies from motivational interviewing, and can query and visualize a user’s health data from a wearable through tool use.
Matthew Jörke tweet media
English
1
1
6
1.3K
Matthew Jörke
Matthew Jörke@mjoerke·
Through formative interviews with 22 participants, we learned that *all* health experts adopted a facilitative approach that did not give unsolicited advice. Notably, this contrasts with how current LLMs are trained to answer questions and give advice.
English
1
0
4
418
Matthew Jörke
Matthew Jörke@mjoerke·
I'm excited to share our new paper exploring how LLM chatbots can be used to motivate physical activity behavior change! Check out our pre-print on arXiv: arxiv.org/abs/2405.06061
Matthew Jörke tweet media
English
1
19
49
10.3K