Sadra Sabouri

PhD Candidate @CSatUSC | Ex-Intern @Microsoft

[Xinyi Zhou @XinyiZ84786]

🧑‍💻👩‍💻Programming is shifting from generating to evaluating. But are we evaluating AI fairly?

Our #ICSE2026 paper showed 56.4% of LLM interactions involved at least one bias!

https://t.co/q8V6tCUVDX

Huge thanks to Zeinabsadat Saghi, @sadrasabourih, and my advisor @AKA_Rini!

[Sadra Sabouri @sadrasabourih]

@nicochristie Would love to test this

[Alex Spangher @AlexanderSpangh]

Neurips 2025 was such a blast! We snuck a grand piano into the CreativeAI Track to demo Aria, our pretrained chat-style music model:

[Sadra Sabouri @sadrasabourih]

@JohnHolbein1 One could argue that significant results are more likely to be published; hence, there's a cut in the graph due to sampling bias (the published papers versus all the research being done).

[Run Huang @runhuang_]

Ever wondered what your fellow researchers think of a paper you're reading? What if you could surf through peer insights without drowning in social media chaos?

Introducing SURF🏄, an interface to bring social media discussions into the paper-reading experience!
#UIST2025🧵(1/n)

[Sadra Sabouri @sadrasabourih]

@Anu_G91 Thank you, I'd be happy to join the fest; but I'm not Indian.

[Sadra Sabouri @sadrasabourih]

🚨 New paper at #VLHCC25 🚨
📄 Exploring the Challenges and Opportunities of AI-assisted Codebase Generation

What are the challenges software engineers face with Codebase Agents (like as Cursor AI)?
arxiv: https://t.co/71Tm2UNWN0
#AI #SoftwareEngineering #Programming #vibecoding

[Philipp Eibl @ominous_zebra]

We conducted a user study with 16 participants—software engineers from both industry and academia—while they completed 3 programming tasks using GitHub Copilot and GPT-Engineer. The tasks varied in length and type, and we measured satisfaction with the generated code.

[Sadra Sabouri @sadrasabourih]

@ominous_zebra @lovable @GptEngineer @GitHubCopilot

[Philipp Eibl @ominous_zebra]

What devs actually want from code AI? Our recently accepted paper at #VLHCC25 digs into it.🧵

📷 Paper: https://t.co/GtLkUpLdT5

[Brihi Joshi @BrihiJ]

Our poster slot got moved, so I'll be talking more about this work and in general about personalizing natural language explanations on Tuesday, July 29th at 4PM in Hall X4 X5 at @aclmeeting!

In case you miss it, our poster is here 👇🏽

[Jonathan K. Kummerfeld @jkkummerfeld]

I'll also be at ACL presenting this, along with 2 papers on clinical coding from my group!

[Joy Wongkamjan @joywwong]

Our paper CTRL-D is accepted to ACL Findings and will be presented at ACL 2025!

🗓️Poster session: 18:00–19:30 (Level 0 Exhibit Halls X4/X5)

I’m sad I can’t be there, but Jordan (@boydgraber) will! You’ll enjoy learning about CTRL-D from him.

Now… what is CTRL-D? 🔍

[Johnny Tian-Zheng Wei @johntzwei]

Are you a researcher, trying to build a small GPU cluster? Did you already build one, and it sucks?

I manage USC NLP’s GPU cluster and I’m happy to offer my expertise. I hope I can save you some headaches and make some friends. Please reach out!

[justin c @HJCH0]

Can you tell what actions are being mimed in this video? If so, you’re smarter than AI models! Check the last tweet in this thread for answers.

In a new paper, we present MIME, which evaluates whether vision language models (VLMs) have a robust understanding of human actions.
🧵

[Brihi Joshi @BrihiJ]

Can LLMs appropriately explain “Why is the sky blue?”
…to a 10-year-old 👶🏽 vs. someone with a PhD in physics 👩🏽‍🔬?

In our #ACL2025 paper, we evaluate how well how well LLMs can tailor their explanations to different people.✍️

🔗 https://t.co/AvP13zNMjY
🧵 (1/n)

[Sadra Sabouri @sadrasabourih]

🛠️ From all this, we developed four guidelines for a more trustworthy relationship between developers and AI.

Check out that and more details from our paper at: https://t.co/OKoDazrXKb

[Sadra Sabouri @sadrasabourih]

🚀 Excited to share that we'll be presenting our paper at #ICSE2025 in Ottawa next week! 🇨🇦
📄 "Trust Dynamics in AI-Assisted Development: Definitions, Factors, and Implications"

Huge thanks to my collaborators @nmedvidovic, @LarsLindemann2, and my supervisor @AKA_Rini! 🙌

[Sadra Sabouri @sadrasabourih]

🤔 RQ3 findings:
Trust is not static — developers often change their minds!
+ They misjudge correctness and later catch bugs
+ They accept changes that introduce unwanted side effects
#AITrust #SoftwareEngineering