iiinova — where your digital self keeps evolving

Come join our group! Still one day left for applying. 😊

000

3) Aligning LLMs on political opinions with English data transfers to all other Western languages we've evaluated on: FR, IT, GE, ES. Congrats team for the acceptance and for the great work! @franziweeber.bsky.social will be presenting it in person at EACL between March 24–29. 😊

Models do become more right-leaning on close-ended questions, but they only become a little less left-leaning on open-ended evaluations such as writing opinionated paragraphs on certain political issues.

100

2) Aligning LLMs with DPO on right-leaning opinions does have an impact on the stance of the models. However, this comes with a caveat.

100

Some findings that I find particularly impactful for the area of political biases in LLMs: 1) Aligning LLMs with DPO on left-leaning opinions does not have a significant impact on the stance of the models given that vanilla LLMs already reflect a more left-leaning alignment.

MilaNLP Lab @milanlp.bsky.social · 18/12/2025

Reposted by Tanise Ceron

🚀 We’re opening 2 fully funded postdoc positions in #NLP! Join the MilaNLP team and contribute to our upcoming research projects. 🔗 More details: milanlproc.github.io/open_positio... ⏰ Deadline: Jan 31, 2026

01913

Tanise Ceron @taniseceron.bsky.social · 02/12/2025

I will be @euripsconf.bsky.social this week to present our paper as non-archival at the PAIG workshop (Beyong Regulation: Private Governance & Oversight Mechanisms for AI). Very much looking forward to the discussions! If you are at #EurIPS and want to chat about LLM's training data. Reach out!

094

Tanise Ceron @taniseceron.bsky.social · 27/11/2025

We could fool ourselves saying that it's because there's no panettone in other periods of the year :P

Tanise Ceron @taniseceron.bsky.social · 27/11/2025

We go out of the routine every now and then at the lab. :)

Tanise Ceron @taniseceron.bsky.social · 24/11/2025

In this paper, we investigate how well media frames generalize across different media landscapes. The 15 MFC frames remain broadly applicable, but requires revisions of the guidelines to adapt to the local context. More on aclanthology.org/2025.starsem...

aclanthology.org

041

Tanise Ceron @taniseceron.bsky.social · 24/11/2025

@agnesedaff.bsky.social presented our work on "Generalizability of Media Frames: Corpus creation and analysis across countries" at *SEM co-located with EMNLP 2025 in China.

291

Tanise Ceron @taniseceron.bsky.social · 18/11/2025

Does anyone know any good resource that systematically documents information about the training data of different LLMs (e.g. name of datasets, language proportion, etc whenever available)?

220

Reposted by Tanise Ceron

MilaNLP Lab @milanlp.bsky.social · 31/10/2025

Proud to present our #EMNLP2025 papers! Catch our team across Main, Findings, Workshops & Demos 👇

12114

Tanise Ceron @taniseceron.bsky.social · 16/10/2025

As I wasn't at the conference, I'd love to be able to watch the recording. Is it available online anywhere? :)

100

- Finally, and for me, most interestingly, our analysis suggests that political biases are already encoded during the pre-training stage. Taken these evidences together, we highlight important implications these results play on data processing in the development of fairer LLMs.

- There's a strong correlation (Pearson r=0.90) between the predominant stances in the training data and the models’ behavior when probed for political bias on eight policy issues (e.g., environmental protection, migration, etc).

120

- Source domains of pre-training documents differ significantly, with right-leaning content containing twice as many blog posts and left-leaning content 3 times as many news outlets.

- The framing of political topics varies considerably: right-leaning labeled documents prioritize stability, sovereignty, and cautious reform via technology or deregulation, while left-leaning documents emphasize urgent, science-led mobilization for systemic transformation and equity.

- left-leaning documents consistently outnumber right-leaning ones by a factor of 3 to 12 across training datasets. - pre-training corpora contains about 4 times more politically engaged content than post-training data.

We have the answers of these questions here : arxiv.org/pdf/2509.22367 We analyze the political content of the training data from OLMO2, the largest fully open-source model. 🕵️‍♀️ We run an analysis in all the datasets (2 pre- and 2 post-training) used to train the models. Here are our findings:

arxiv.org

arXiv cs.CL Computation and Language @cscl-bot.bsky.social · 29/09/2025

📣 New Preprint! Have you ever wondered what the political content in LLM's training data is? What are the political opinions expressed? What is the proportion of left- vs right-leaning documents in the pre- and post-training data? Do they correlate with the political biases reflected in models?

24613

Reposted by Tanise Ceron

Tanise Ceron, Dmitry Nikolaev, Dominik Stammbach, Debora Nozza: What Is The Political Content in LLMs' Pre- and Post-Training Data? arxiv.org/abs/2509.22367 arxiv.org/pdf/2509.22367 arxiv.org/html/2509.22367

013

Thanks SoftwareCampus for supporting Multiview, the organizers of INRA, and Sourabh Dattawad and @agnesedaff.bsky.social for the great collaboration!

Our evaluation with normative metrics shows that this approach does not diversify only frames in user's history, but also sentiment and news categories. These findings demonstrate that framing acts as a control lever for enhancing normative diversity.

100

In this paper, we propose introduce media frames as a device for diversifying perspectives in news recommenders. Our results show an improvement in exposure to previously unclicked frames up to 50%.

100

Joachim Baumann @joachimbaumann.bsky.social · 12/09/2025

Today Sourabh Dattawad presented our work "Leveraging Media Frames to Improve Normative Diversity in News Recommendations" at INRA (International Workshop on News Recommendation and Analytics) co-located with RecSys 2025 in Prague. arxiv.org/pdf/2509.02266

121

Reposted by Tanise Ceron

🚨 New paper alert 🚨 Using LLMs as data annotators, you can produce any scientific result you want. We call this **LLM Hacking**. Paper: arxiv.org/pdf/2509.08825

$We present our new preprint titled "Large Language Model Hacking: Quantifying the Hidden Risks of Using LLMs for Text Annotation". We quantify LLM hacking risk through systematic replication of 37 diverse computational social science annotation tasks. For these tasks, we use a combined set of 2,361 realistic hypotheses that researchers might test using these annotations. Then, we collect 13 million LLM annotations across plausible LLM configurations. These annotations feed into 1.4 million regressions testing the hypotheses. For a hypothesis with no true effect (ground truth $p > 0.05$), different LLM configurations yield conflicting conclusions. Checkmarks indicate correct statistical conclusions matching ground truth; crosses indicate LLM hacking -- incorrect conclusions due to annotation errors. Across all experiments, LLM hacking occurs in 31-50\% of cases even with highly capable models. Since minor configuration changes can flip scientific conclusions, from correct to incorrect, LLM hacking can be exploited to present anything as statistically significant.$

6303106

Reposted by Tanise Ceron

MilaNLP Lab @milanlp.bsky.social · 07/07/2025

Last week we held our 1st MilaNLP retreat by beautiful Lago Maggiore! ⛰️🌊 We shared research ideas, stories (academic & beyond), and amazing food. It was a great time to connect outside of the usual lab working days, and most importantly, strengthen our bonds as a team. #ResearchLife #NLProc

0225

Reposted by Tanise Ceron

Beatrice Savoldi @bsavoldi.bsky.social · 03/06/2025

🔍 Stiamo studiando come l'AI viene usata in Italia e per farlo abbiamo costruito un sondaggio! 👉 bit.ly/sondaggio_ai... (è anonimo, richiede ~10 minuti, e se partecipi o lo fai girare ci aiuti un sacco🙏) Ci interessa anche raggiungere persone che non si occupano e non sono esperte di AI!

bit.ly

Qualtrics Survey | Qualtrics Experience Management

The most powerful, simple and trusted way to gather experience data. Start your journey to experience management and try a free account today.

11618

Tanise Ceron @taniseceron.bsky.social · 15/05/2025

Reminder for the importance of evaluating political biases robustly. :)

Dirk Hovy @dirkhovy.bsky.social · 03/05/2025

Reposted by Tanise Ceron

We (w/ @diyiyang.bsky.social, @zhuhao.me, & Bodhisattwa Prasad Majumder) are excited to present our #NAACL25 tutorial on Social Intelligence in the Age of LLMs! It will highlight long-standing and emerging challenges of AI interacting w humans, society & the world. ⏰ May 3, 2:00pm-5:30pm Room Pecos

0156

Reposted by Tanise Ceron

Verena Kunz @verenakunz.bsky.social · 30/04/2025

Join us in an hour at 17:00 (CEST) for @taniseceron.bsky.social's talk on "Evaluating Political Bias: Insights into Robustness and Multilinguality“. Access to Zoom at join.slack.com/t/tadapolisc... or send me a ✉️

011

Tanise Ceron @taniseceron.bsky.social · 23/04/2025

Sure, it's here: github.com/tceron/eval_... The code mapping is in the readme file. :)

github.com

Christopher Klamm 🏔️ @cklamm.bsky.social · 21/04/2025

Reposted by Tanise Ceron

🥁 It's the second half of our 🌱 speaker series (tada.cool) this term, and we couldn't be more excited! Next week (Wednesday, April 30 at 5pm CET), we have the pleasure of welcoming @taniseceron.bsky.social to share insights on "Facilitating Information Access Through Language Models". More details ⬇️

075

Tanise Ceron @taniseceron.bsky.social · 22/04/2025

change the political worldviews of models. In our study, we find that the previous version (Llama-2) consistently reflects more left-leaning views. However, it does depend on the policy issue as we found clear stances of the models only towards social state welfare, environment protection and [2/3]

Tanise Ceron @taniseceron.bsky.social · 22/04/2025

I agree 100% that we need to understand what they're measuring, and specifically, how they're aligning the models to be hold certain types of political worldviews. However, I find your results rather puzzling because Llama3.1 was released much before they started announcing their strategy to [1/3]

170

Tanise Ceron @taniseceron.bsky.social · 25/03/2025

All the very best for this new chapter @florplaza.bsky.social! 😃 We already miss you here! ❤️

Dirk Hovy @dirkhovy.bsky.social · 05/03/2025

Reposted by Tanise Ceron

Wanna keep up with our @milanlp.bsky.social lab? Here is a starter pack of current and former members: bsky.app/starter-pack...

0137

Tanise Ceron @taniseceron.bsky.social · 05/03/2025

Happy to be presenting at #TaDa and looking forward to watching the great talks coming up. :)

032

Tanise Ceron @taniseceron.bsky.social · 17/02/2025

Hmm, I agree that this could be a good solution. Though I wonder if this is feasible based on the pace that advancements take place in this area.

020

Tanise Ceron @taniseceron.bsky.social · 17/02/2025

That could def encourage people to polish more, but I think we need more well-defined categories. E.g. paper with best related work section given that this section is often underestimated nowadays and it's an important step to build on people's previous work. Ofc, this is just one among many! :)

Tanise Ceron @taniseceron.bsky.social · 17/02/2025

It decreases the burden of reviewers whose role may not be to give feedback, but to check the quality and validity of a given research piece. Curious to know what other researchers think about it. :)