opportunities

Are we recruiting?

Yes, we are actively recruiting PhD students in this 🍁 Fall! Please see below to figure out if your research interests align with ours, and apply to the USC PhD Admissions by ⚡ Dec 15th, 2023 and select Swabha as a potential advisor. USC has waived GRE scores for graduate admissions and offers fee waivers to select applicants.

What do we work on?

Here are some questions we have worked on recently:

🖋️ What do we understand about the generative capabilities of language models?

Has language model generation reached a performance saturation or do language models still make systematic errors owing to their design? We studied the softmax bottleneck as one concrete limitation and how that affects language generation and justifies truncation sampling. Can we build better language generators?

🎨 Can language models handle all types of language?

How do language models handle specific distributions of language, such as ambiguous language, comparative language or the language of explanations? Can language models generate structured data?

🕵🏼‍♀️ Are we evaluating the generative capabilities of models rigorously?

What cannot be measured, cannot be improved. Can we reliably compare the generative performance of two different models, in either close-ended generation tasks such as summarization or in open-ended generation? What makes model A better than model B, or are our test sets somewhat misleading us?

📚 What does our data tell us about our models?

What makes a data collection valuable for instruction tuning or finetuning large language models? Is all human feedback equally valuable under PPO or DPO? Our Dataset Cartography offers point estimates, and V-Information offers both point and aggregate estimates of data quality. How can we build similar estimates for generative models?

Are all modalities and all data necessary in multimodal settings?

🤖 What can our models reveal about our society?

How far can language models go in helping us understand complex social phenomena such as homelessness? Is it possible to create collaborative setups between humans and generative models to this end? What role does conversational and social context play in this understanding? Can socio-technical solutions work well for all?

Should you send Swabha an email?

Due to the high volume of email, there is very little chance Swabha will be responsive to your email, sorry! See special cases below.  

PhD Applicants

Due to the high volume of email, there is very little chance we will be able to respond, sorry! However, we will carefully consider every single PhD application we receive.

If you’re already at USC:

• PhD Students: If you think your interests overlap considerably with DILL and would like to collaborate, do email Swabha.
• Masters Students:
  • RAships: We might recruit motivated Masters students with research experience based on availability of time in Spring 2024. If interested in our lab, please email Swabha a copy of your CV and/or a link to your website.
  • TAships: We are not looking for any Masters Students as TAs at the moment.
  • Undergrads: We recruit undergrad researchers almost exclusively through the CURVE program at USC. We do NOT have any openings for new undergrads in Spring 2024.

If you’re not at USC:

• High-school students: Please consider K-12 research opportunities at USC.
• Prospective Masters students: At the moment, we have no openings for you - apologies!
• Prospective PhD students: See above.
• Visiting PhD students: If you think your interests overlap considerably with DILL and would like to collaborate, do email Swabha.
• Visiting Undergrads: At the moment, we have no openings for you - apologies!
• Others: Swabha might NOT respond if you’re looking for recruitment - apologies! Other emails about research questions etc. are welcome, though she might be slow to respond to them.