building sth new … | PhD | YC25 | prev. Mercura, Meta, ADSK & Harvard
Munich
Joined January 2017
- Tweets73
- Following303
- Followers183
- Likes343
It’s becoming increasingly difficult to make sense of claude outputs. I often have to prompt it to rewrite the output in simple terms and remove all of all the fluff. Is this a gimmick to make us burn more tokens?
H.Dang retweeted
#ICML2026 Come check out our poster today on training LLMs for creative and open-ended tasks using stateful user simulators!
10:30 AM | Hall A, Poster #801
Ask an LLM for a "post that'll pop off". Output misses so you say "more unique." It asks "unique how?" but you don't know yet.
DiscoverLLM (ICML'26): trains LLMs to help users discover their intents, not just execute them.
📑 arxiv.org/abs/2602.03429
🌐 taesookim.com/discoverllm
H.Dang retweeted
This is Ditto.
It clones any website in less than 5 minutes.
Clean code w/ components, tokens, interactions, hover effects.
Free & open source. REST API and MCP server included.
H.Dang retweeted
Handheld camera motion for Seedance 2.0
I hate blender so I stage it on my iphone.
You stage the scene (open source link below) and can walk around it and capture real camera movement in 3D space, then Seedance populates the scene.
Handheld camera motion and realistic shake is hard to describe or prompt to AI. The iphone’s ARKit is super precise with motion tracking and you can record complex scenes as a one man studio.
Here is the AI filmmaking workflow:
1. Set up blocking in the 3D space
2. Record camera motion
3. Generate a starting image with location and characters
4. Feed image and video to Seedance
This was a fun build with claude code. Happy if others want to build on it. – at Palo Alto, CA
Seedance 2.0 is insane.
Blender previs is having a moment in AI filmmaking for a reason.
I wanted to see how far I could push the workflow beyond a simple camera move in an ambush scene.
The workflow:
1. Make a start image in Midjourney
2. Block it out in Blender, animate the camera
3. Feed both to Seedance
The more I test this pipeline, the more convinced I am that hybrid pipelines are the future. The conversation is moving fast but previs is one of the clearest unlocks so far.
Seedance still isn't perfect at matching the reference but it's close.
You can experiment with staging, timing, and camera movement before you ever touch the video model.
Solve the scene before you generate the shot.
Learned that friend burns at least as many tokens for running e2e agent test&evals as he used for coding new features.
H.Dang retweeted
I’ll be presenting EvalLM tomorrow 9:15am (Tue) at #CHI2024 in the “Evaluating AI Technologies B” (Room 320). You can also find the link to the live demo, code and paper in our website: evallm.kixlab.org
Improving LLM outputs is difficult because evaluating these outputs is difficult...
Introducing EvalLM ⚗️ - an interactive system for evaluating LLM outputs on user-defined and application-specific criteria 📏
🔗: evallm.kixlab.org
📄: arxiv.org/abs/2309.13633
(1/n)
H.Dang retweeted
Now available in the ACM DL: doi.org/10.1145/3613904.3642…
If you're at #CHI2024, the talk is at Wed 2PM in room 310! Or find us at a coffee machine near you. 😄 @sigchi
✍️Compelling writing is tailored to an audience but it's hard to ask them while writing. Can #LLMs help? 🤔
👉We explore this as "Writer-Defined #AI Personas for On-Demand Feedback Generation"
#CHI2024 preprint: arxiv.org/abs/2309.10433
@KarimBenharrak @flolehmann_de @hen_drik
ALT Illustration with emojis. A "writer" emoji is thinking about a set of four other faces, that is, people. Text next to it says: "Can text generation support writers in considering reader perspective?"
H.Dang retweeted
💡Ever missed a breathtaking view while driving? Tired of juggling apps to perform digital actions? #CHI2024
Meet OmniActions: an egocentric, multimodal AI agent in #PervasiveAR that predicts your next move during real-world interaction.
Preprint: arxiv.org/abs/2405.03901
How can we enable more flexible interactions with computers using AI?
🌴 Join us for lunch on Monday, May 13th, at #CHI2024 🌊 if you are excited about topics surrounding abstraction ladders🪜 , dynamic notations and adaptive interfaces.
Info: dynamicabstractions.github.i…
Want to reimagine how we interact with computers in the age of AI?
And excited about ... 🪜 #abstraction_ladder, 🌊 #dynamic_medium, 🔣 #notations, and 🔀 #diagrams?
🌴 Join us for Monday (05/13) lunch at #CHI2024!
shorturl.at/noxU1
🚀 Learn more:
dynamicabstractions.github.i…
H.Dang retweeted
<Summer Research Internship at @kixlab_kaist>
I'm looking for undergrad research interns to join my research group (kixlab.org) this summer. Most projects this round are about human-AI interaction. Please share broadly!
juhokim.com/2024-summer-inte…
H.Dang retweeted
We're hiring!
The HCI & Visualization group at Autodesk Research is hiring for two Research Scientist positions at the intersection of HCI and AI.
If interested, please apply (and reach out if you have questions).
autodesk.wd1.myworkdayjobs.c…
(@ADSKResearch)
H.Dang retweeted
Looking to do a PhD in HCI? Going to #CHI2024? Look no further! The Montréal HCI lab is recruiting PhD students (hci.iro.umontreal.ca). I will be at CHI this year, reach out if you’d like to chat! 🤩✨
H.Dang retweeted
Unfortunately, Large Language Models will not consistently follow the instructions that you give them. This is a massive problem when you are building AI systems that require a particular type of output from the previous step to feed into the next one!
For example, imagine you are building a blog post writing system that first takes a question and retrieved context to output a list of topics. These topics have to be formatted in a particular way, such as a comma-separated list or a JSON of Topic objects, such that the system can continue writing the blog post! 🤝
I am SUPER excited to share the 4th video in my DSPy series, diving into 3 solutions to structuring outputs in DSPy programs: (1) TypedPredictors, (2) DSPy Assertions, and (3) Custom Guardrails with the DSPy programming model! 📖
TypedPredictors follow the line of thinking around JSON mode and using Pydantic BaseModels to interface types and custom objects into a JSON template for LLMs. The output can then be validated to provide a more structured retry prompt to correct the output structure! 🛠️
DSPy Assertions are one of the core building blocks of DSPy, offering an interface to input a boolean-valued function and a retry prompt which is templated alongside the past output to retry the call to the LLM! ⚠️♻️
Custom Guardrails with the DSPy Programming Model are one of the things I love the most about DSPy — we have unlimited flexibility to control these systems however we want. The video will also show you how to write custom guardrails and retry Signatures and discussion around using TypedPredictors for your Custom Guardrails and potentially feeding your Custom Guardrails into a DSPy Assertion. 💥
I had so much fun exploring this topic! Further seeing how well OpenAI’s GPT-4 and GPT-3.5-Turbo, Cohere’s Command R, and Mistral 7B hosted with Ollama perform with each Structured Output strategy! I also found monitoring structured output retries to be another fantastic application of Arize Phoenix! I hope you find the video useful! 🎙️
youtube.com/watch?v=tVw3CwrN…
H.Dang retweeted
I’m looking for HCI graduate students to join my research group for an in-person internship this summer! ☀️
More details on how to apply ⬇️
H.Dang retweeted
Please RT! We've developed an AI-assisted tool for planning and making videos, and we're looking for expert video creators to try it out and provide feedback. We're especially looking for creators experienced in making explainer vidoes. Sign up here! tinyurl.com/8s7b5xtx
H.Dang retweeted
Join us at the ELLIS Summer School on Collaborative and Generative AI in Helsinki, Finland, from July 1st to 5th, 2024!
For more information, or to register, see fcai.fi/ellis-summer-school-…
#AI #CoGenAI #ELLIS #SummerSchool #Helsinki #AaltoUniversity
H.Dang retweeted
Please share: We are looking for a new PhD student for the research project “Human-Computer Interaction in Extended Reality (XR)” – read more here: phd.nat.au.dk/for-applicants…