Laia Balcells
LightNews LightNews
banner
laiabalcells.bsky.social
Laia Balcells
@laiabalcells.bsky.social
7.2K followers 1.5K following 790 posts

Yale educated political scientist. Georgetown professor. Reskeets are not endorsements.

Videos
Political science 60%
Sociology 17%
Posts Replies Media Videos

Reposted by Laia Balcells, Zen Faulkes

ketanjoshi.co
Ketan Joshi @ketanjoshi.co · 2d
Deeply absurd. This Google PDF published on a blog (arxiv, not peer reviewed) claims an LLM is "PhD level" but in most cases the MAJORITY of reference URLs were invalid or inaccessible.

A PhD sitting down and just fabricating >50% of sources = career ending

arxiv.org/abs/2511.11597
CLINB: A Climate Intelligence Benchmark for Foundational Models Michelle Chen Huebscher1 , Katharine Mach2 , Aleksandar Stanić1 , Markus Leippold1,3, Ben Gaiarin1 , Zeke Hausfather4 , Elisa Rawat , Erich Fischer5 , Massimiliano Ciaramita1 , Joeri Rogelj6 , Christian Buck1 , Lierni Sestorain Saralegui1 and Reto Knutti5 1Google DeepMind, 2University of Miami, 3University of Zurich, 4Stripe, 5ETH Zurich, 6 Imperial College London Evaluating how Large Language Models (LLMs) handle complex, specialized knowledge remains a critical challenge. We address this through the lens of climate change by introducing CLINB, a benchmark that assesses models on open-ended, grounded, multimodal question answering tasks with clear requirements for knowledge quality and evidential support. CLINB relies on a dataset of real users’ questions and evaluation rubrics curated by leading climate scientists. We implement and validate a model-based evaluation process and evaluate several frontier models. Our findings reveal a critical dichotomy. Frontier models demonstrate remarkable knowledge synthesis capabilities, often exhibiting PhD-level understanding and presentation quality. They outperform “hybrid" answers curated by domain experts assisted by weaker models. However, this performance is countered by failures in grounding. The quality of evidence varies, with substantial hallucination rates for references and images. We argue that bridging this gap between knowledge synthesis and verifiable attribution is essential for the deployment of AI in scientific workflows and that reliable, interpretable benchmarks like CLINB are needed to progress towards building trustworthy AI systems. Total Reference URLs Generated claude-opus-4-1 claude-sonnet-4 gpt-5 hybrid gemini-2.5-pro gemini-2.5-flash o3 0.0 0.2 0.4 0.6 0.8 1.0 Proportion Reference URL Status hybrid gemini-2.5-pro claude-opus-4-1 o3 gemini-2.5-flash claude-sonnet-4 0 200 400 600 800 1000 Count of URLs Total Image URLs Generated hybrid claude-opus-4-1 gemini-2.5-flash gemini-2.5-pro claude-sonnet-4 o3 0.0 0.2 0.4 0.6 0.8 1.0 Proportion Image URL Status Status OK INACCESSIBLE_CONTENT INVALID_URL ERROR Figure 3 | Number of reference (top), and image (bottom), URLs and their status. Ablations We perform several ablation studies with the autorater (Table 4). Notably, removing the question-specific rubrics from the prompt changes the results only in the bottom half, with the Hybrid answers overtaken by Gemini 2.5 Flash and Claude Sonnet 4. This suggests that the additional resolution provided by the rubrics applies primarily to the kind of responses used to develop the rubrics. Or, in other words, that rubrics are far from complete. Hence, it is important that rubrics adapt to new data as better models become availab A New Expert-Grounded Benchmark for Scientific AI We introduce CLINB, a benchmark for modelbased evaluation of frontier models on complex, multimodal scientific communication. Its core is a new dataset of real-world climate questions paired with data-driven, question-specific evaluation rubrics, curated and validated by leading climate scientists through a novel three-phase, human-in-the-loop process.2 PhD-Level Synthesis vs. Attribution Failures Frontier models demonstrate remarkable knowledge synthesis, often exhibiting a PhD-level understanding. However, this performance masks a critical inadequacy in grounding. We report substantial hallucination rates for references (10% to 25%) and even more failures for images (50% to 80% in certain settings), exposing a major gap between synthesis and verifiable attribution. Insights into Human-AI Collaboration Dynamics Autonomous frontier models surpass ’hybrid’ answers (curated by experts using weaker AI assistance), revealing the assisting model’s capability—not human oversight—as the primary bottleneck. Counter-intuitively, highly motivated non-specialists (our ’Advocates’) who deeply engage with AI tools can produce higher-quality answers than domain experts who engage less with AI during answer curation. A Validated Methodology for Scalable Oversight We validate a rigorous, rubric-based autorater. Ablation studies demonstrate that structured prompts and automated evidence-checking are essential for mitigating inherent LLM judge biases. This process is hampered by inaccessible sources (up to 50%). Furthermore, we identify evaluation challenges, including model familiarity bias in human raters and the limitations of rubrics to generalize across models.

laiabalcells.bsky.social
Laia Balcells @laiabalcells.bsky.social · 2d
I was very puzzled when I saw this. Time to unsubcribe from The Atlantic?
nycsouthpaw.bsky.social southpaw @nycsouthpaw.bsky.social · 2d
This cover oozes with contempt for the magazine’s history and its readership
theatlantic.com The Atlantic @theatlantic.com · 2d
Over many interviews, Robert F. Kennedy Jr. told Michael Scherer about how he plans to remake America’s public-health system. Can he lead the scientific establishment he’s spent much of his life crusading against?

Read more in our new cover story: theatln.tc/q9hXxlbM

Reposted by Laia Balcells

nycsouthpaw.bsky.social
southpaw @nycsouthpaw.bsky.social · 2d
This cover oozes with contempt for the magazine’s history and its readership
theatlantic.com The Atlantic @theatlantic.com · 2d
Over many interviews, Robert F. Kennedy Jr. told Michael Scherer about how he plans to remake America’s public-health system. Can he lead the scientific establishment he’s spent much of his life crusading against?

Read more in our new cover story: theatln.tc/q9hXxlbM

Reposted by Laia Balcells, Josh Pasek, David Darmofal

jenniferhoewe.bsky.social
Jen Hoewe @jenniferhoewe.bsky.social · 2d
The Center for American Political History, Media, and Technology (CAPT) at Purdue University is currently searching for a Postdoctoral Research Associate! The deadline to apply is Dec. 31, 2025. More information about the position is available here: careers.purdue.edu/job/Postdoct...
#commsky
Postdoctoral Research Associate
Postdoctoral Research Associate
careers.purdue.edu

laiabalcells.bsky.social
Laia Balcells @laiabalcells.bsky.social · 2d
But Spain is the EU country with greatest economic growth, etc.

Reposted by Laia Balcells

eldiario.es
elDiario.es @eldiario.es · 2d
‼️ÚLTIMA HORA | Telefónica plantea a los sindicatos un ERE de 5.000 trabajadores, un 20% de la plantilla
Telefónica plantea a los sindicatos un ERE de 5.000 trabajadores, un 20% de la plantilla
La operadora comunica a los sindicatos el alcance de un ajuste que afecta a sus divisiones clave en España: Telefónica Móviles, Telefónica Móviles y Telefónica Soluciones
www.eldiario.es

laiabalcells.bsky.social
Laia Balcells @laiabalcells.bsky.social · 3d
Brett Goldstein (Ted Lasso) is very funny. Does anyone have the clip of his joke about US bathrooms (from the show The Second Best Night of your Life)? It's hilarious.

Reposted by David Dunning, Laia Balcells, Scott Lucas

atrupar.com
Aaron Rupar @atrupar.com · 3d
WELKER: Inflation has gone up. It's at 3% now up from 2% in April when the tariffs were imposed--

BESSENT: No, no no no. So, inflation hasn't gone up. The one thing we're not gonna do is do what the Biden administration did and tell the American people they don't know how they feel

laiabalcells.bsky.social
Laia Balcells @laiabalcells.bsky.social · 3d
How I recovered from a collaboration gone wrong
www.science.org/content/arti...
How I recovered from a collaboration gone wrong
After learning “how quickly ‘we’ can turn wicked,” this researcher reshaped his approach
www.science.org

laiabalcells.bsky.social
Laia Balcells @laiabalcells.bsky.social · 3d
The country/world we live in now.
atrupar.com Aaron Rupar @atrupar.com · 3d
TAPPER: Right before the election, Trump threatened to withhold federal funding from New York City if Mamdani were elected. Was that just bluster? He doesn't mean it?

HASSETT: Well, it feels like he doesn't mean it now

Reposted by Laia Balcells, Mark Rice

atrupar.com
Aaron Rupar @atrupar.com · 3d
TAPPER: Right before the election, Trump threatened to withhold federal funding from New York City if Mamdani were elected. Was that just bluster? He doesn't mean it?

HASSETT: Well, it feels like he doesn't mean it now

Reposted by Laia Balcells

hofrench.bsky.social
Howard French @hofrench.bsky.social · 3d
Quietly terrifying. www.wsj.com/economy/aida...
An MIT Student Awed Top Economists With His AI Study—Then It All Fell Apart
Aidan Toner-Rodgers shot to academic fame in a field hungry for new insights and revelatory research. But a computer scientist thought something seemed off.
www.wsj.com

Reposted by Juanjo Medina

laiabalcells.bsky.social
Laia Balcells @laiabalcells.bsky.social · 3d
OPINIÓN | Abusos en la universidad: por qué persiste la impunidad y qué reformas pueden cambiarla, por Aina Gallego www.eldiario.es/132_c306cd?u...
Abusos en la universidad: por qué persiste la impunidad y qué reformas pueden cambiarla
Las universidades españolas arrastran casos de acoso laboral y sexual que rara vez se sancionan, en un ecosistema marcado por la protección reputacional y la endogamia. Romper estas dinámicas con movi...
www.eldiario.es

Reposted by Laia Balcells

kityates.bsky.social
Kit Yates @kityates.bsky.social · 3d
"Boris Johnson took four days off from official government business during a key period in the UK’s Covid preparation when the NHS was bracing to be “overwhelmed” by the virus."

When we most needed a leader we, instead, had this guy.
Boris Johnson took four days off as NHS warned Covid could ‘overwhelm’ system
Files show then PM was walking dog, riding motorbike and hosting guests as pandemic planning stalled in ‘lost month’
www.theguardian.com

Reposted by Rosemary A. Joyce, Laia Balcells, Laurent Dubois , and 2 more

bryguy.bsky.social
bryan! @bryguy.bsky.social · 5d
This might be the single funniest thing that has happened during this incredibly cursed year

Reposted by Steve Peers, Rosemary A. Joyce, Laia Balcells , and 1 more

kevjg.bsky.social
Kevin @kevjg.bsky.social · 5d
he never looks at JD that way

Reposted by Laia Balcells, Timothy Burke

sethcotlar.bsky.social
Seth Cotlar @sethcotlar.bsky.social · 5d
Ross Douthat: Don’t Worry. Trump Doesn’t REALLY Want Democratic Officials to be Executed.
OPINION There Will Be No Trump Coup A final pre-election case for understanding the president as a noisy weakling, not a budding autocrat. Oct. 10, 2020 Oliver Contreras for The New York Times By Ross Douthat Opinion Columnist

Reposted by Laia Balcells

ibei.org
Institut Barcelona d'Estudis Internacionals (IBEI) @ibei.org · 6d
🎙️ #IBEIopeninglecture of the Academic Year 2025-2026

The Role of Museums in Post-Conflict Societies: A Comparative Assessment. With @laiabalcells.bsky.social, Gallagher Family Professor of Government, Georgetown University

🔎 Check the agenda and join us at the next events: t.ly/qKJLE
#ibei20years

Reposted by Laia Balcells, Dana R. Fisher, David Darmofal

logicallyjc.bsky.social
John Collins @logicallyjc.bsky.social · 6d
This is the way.💯
atrupar.com Aaron Rupar @atrupar.com · 6d
CNN: What if there are more names connected to the Democratic Party that come out in the files?

CHRIS MURPHY: So what?

Reposted by Scott L. Greer, Laia Balcells

darinself.com
Darin Self @darinself.com · 8d
Trendline go (straight) down

Reposted by David G. Rand, Laia Balcells

usapolling.bsky.social
Polling USA @usapolling.bsky.social · 8d
"Do you think that Trump was involved in crimes allegedly committed by Jeffrey Epstein?"

All:
Yes: 42%
No: 34%

Yes Among:
DEM: 79%
IND: 43%
GOP: 7%

YouGov / Nov 17, 2025

Reposted by Laia Balcells

kylegriffin1.bsky.social
Kyle Griffin @kylegriffin1.bsky.social · 8d
BREAKING:

A panel of federal judges has blocked Texas' newly-redrawn congressional map — which made five districts in the state more favorable to Republicans — saying the plan appeared to be an illegal race-based gerrymander.

laiabalcells.bsky.social
Laia Balcells @laiabalcells.bsky.social · 9d
Es Rosalia la primera catalana que surt a un late night show americà?

Reposted by Laia Balcells, Robert B. Reich

ips-dc.org
Institute for Policy Studies @ips-dc.org · 9d
NEW: The Trump administration's brutal removal of so many workers is projected to shrink the nation’s GDP by as much as 6.8% — a deeper hit than the one sustained during the Great Recession.

A.J. Schumann:
Mass Deportations Aren’t Helping Workers. They’re Tanking The Economy. - OtherWords
Trump claimed immigrants were “taking your jobs.” But during his brutal crackdown, the job market has only gotten worse.
otherwords.org

Reposted by Laia Balcells

rbreich.bsky.social
Robert Reich @rbreich.bsky.social · 11d
Trump demanded Texas and other red states redraw their congressional districts to rig the midterms in Republicans' favor.

California fought back by redrawing its map to balance out this power grab. Now, Trump’s DOJ is suing California.

The hypocrisy is endless.

Reposted by Will Jennings, Steve Peers, Laia Balcells , and 8 more

chrismurphyct.bsky.social
Chris Murphy @chrismurphyct.bsky.social · 11d
This is so so disturbing. Trump is now pardoning January 6th rioters for unrelated crimes, just to reward them for their violence to keep him in power.

The Republican Party is in the full time business of endorsing and incentivizing political violence. So scary.

Reposted by Laia Balcells, Olúfẹ́mi Táíwò

maustermuhle.bsky.social
Martin Austermuhle @maustermuhle.bsky.social · 11d
THE WASHINGTON SPIRIT ARE GOING TO THE FINALS!

Reposted by Laia Balcells

javierespinosa.bsky.social
Javier Espinosa @javierespinosa.bsky.social · 11d
Ezio Gavazzeni, el escritor que investigo a los francotiradores turísticos de Sarajevo: "Los clientes (los francotiradores) procedían de muchos países: eran italianos, alemanes, franceses, ingleses, españoles, estadounidenses o canadienses". www.elmundo.es/internaciona...
Ezio Gavazzeni, el escritor tras los francotiradores turísticos: "Había españoles ricos y relevantes en los safaris humanos de Sarajevo"
Las peores barbaries de la guerra en Bosnia no quedarán impunes aunque hayan pasado tres décadas. Con ese espíritu, el escritor Ezio Gavazzeni (Milán, 1959) presentó una...
www.elmundo.es

Reposted by Laia Balcells

filipecampante.bsky.social
Filipe Campante @filipecampante.bsky.social · 11d
This right here is a reason why I think humans will be kept around even in sectors where AI is demonstrably better from a performance standpoint.

Waymo Was on a Roll in San Francisco. Then One of Its Driverless Cars Killed a Cat. www.nytimes.com/2025/11/15/u...

Reposted by Lisa W. Fazio, Will Jennings, Tim Bale , and 6 more