Chara Podimata

We’re launching the LLM Election Observatory: llmelectionobservatory.org, a research project with @adamberinsky.bsky.social, Charles Stewart, Nicolas Emmenegger, Young-Kyung Kim, and Alessandro Morosini on how LLMs respond to election-related questions. @nytimes.com wrote a story about it. 1/n 1/n

I also want to shout out once again the team behind the 2024 election audit (arxiv.org/pdf/2509.18446), from which we drew great inspiration @sarahcen.bsky.social, @aifi.bsky.social, Hedi Driss, @dataspen.bsky.social, and @aleksmadry.bsky.social 8/n

We’re launching the LLM Election Observatory: llmelectionobservatory.org, a research project with @adamberinsky.bsky.social, Charles Stewart, Nicolas Emmenegger, Young-Kyung Kim, and Alessandro Morosini on how LLMs respond to election-related questions. @nytimes.com wrote a story about it. 1/n 1/n

The full dataset will be released publicly on Hugging Face after the election season is complete. If you're interested in learning more, please visit our dashboard: llmelectionobservatory.org, where you can read more about our methodology and even see the exact model responses. 6/n

We’re launching the LLM Election Observatory: llmelectionobservatory.org, a research project with @adamberinsky.bsky.social, Charles Stewart, Nicolas Emmenegger, Young-Kyung Kim, and Alessandro Morosini on how LLMs respond to election-related questions. @nytimes.com wrote a story about it. 1/n 1/n

Each question is also asked as-is and under a set of "framings" (what we call prompt steers) to every model. These prompt steers are designed to generally reflect some ways that people may choose to qualify questions based on their own political standpoints or demographics. 4/n

We’re launching the LLM Election Observatory: llmelectionobservatory.org, a research project with @adamberinsky.bsky.social, Charles Stewart, Nicolas Emmenegger, Young-Kyung Kim, and Alessandro Morosini on how LLMs respond to election-related questions. @nytimes.com wrote a story about it. 1/n 1/n

A bit on our project: We are running a large-scale longitudinal audit of the responses that frontier LLMs are providing to 2026 midterm election-related queries. 2/n

We have a new paper out today, and I’m so proud of it and the work that went into creating it! Link here: arxiv.org/pdf/2509.184.... The 2024 US presidential election was the first major election in the US since the popularization of LLMs. (1/6)

(iii) interrogate the models’ beliefs about candidates’ attributes, and (iv) reveal the models’ implicit predictions of the election outcome. We’re making the dataset publicly available and invite researchers to run their own analyses on it: huggingface.co/datasets/sar.... (5/6)

We have a new paper out today, and I’m so proud of it and the work that went into creating it! Link here: arxiv.org/pdf/2509.184.... The 2024 US presidential election was the first major election in the US since the popularization of LLMs. (1/6)

Our design systematically varies question content and format, resulting in a rich dataset that enables analyses of the models’ behavior over time (e.g., across model updates), sensitivity to steering, responsiveness to instructions, election-related knowledge and “beliefs.” (3/6)

We have a new paper out today, and I’m so proud of it and the work that went into creating it! Link here: arxiv.org/pdf/2509.184.... The 2024 US presidential election was the first major election in the US since the popularization of LLMs. (1/6)