@daniel_sc4

Intern of TS @cohere | PhDing @unimib 🇮🇹 & @GroNlp 🇳🇱, interpretability et similia

danielsc4.it/
Joined February 2015
  • Tweets125
  • Following879
  • Followers248
23Media
You can easily save up to 65% of compute while improving performance on reasoning tasks 🤯 👀 Meet EAGer: We show that monitoring token-level uncertainty lets LLMs allocate compute dynamically - spending MORE on hard problems, LESS on easy ones. 🧵👇
2
5
1
25
6,398
astra/fable and co. may be good at hacking HF and winning prizes, but these models are extremely shit at writing I respect your time, so here’s the solution: style transfer. But I really want to see companies start optimising in this direction in the future
Claude
@claudeaiSep 22
Replying to @claudeai
Opus 5.5 communicates more naturally, addressing some of the most common feedback we heard on Opus 5. It puts the most important information up front and follows the writing rules you give it, which makes long sessions easier to follow.
1
2
89
probably because writing style is incredibly rich and feature dense -> extremely complex to decode. We even tried, it was fun and tough xitter.cf/daniel_sc4/status/1925…
Daniel Scalena@daniel_sc423 May 2025
Replying to @daniel_sc4
📈 But can models recognize and replicate individual translator styles?: ✓ Classifiers can find styles with high acc. (humans kinda don’t) ✓ Multi-shot prompting boosts style a lot ✓ We can detect strong style traces in activations (esp. mid layers) 3/
2
31
agents are taking over swe tasks, math, biology and it just keeps going while we're losing ground on their monitorability (no, cot aint good). mech interp is undoubtedly cool at explaining narrow things about those agents but something in between is missing this is it:
Parallax
@parallx_aiSep 9
Introducing Parallax 🔍 We're a new UK/EU-based nonprofit research lab building scalable methods and infrastructure to audit the beliefs, goals, and plans that shape model behaviours in long-horizon agentic evaluations. parallx.ai — Thread 🧵 1/
1
1
293
for those worried that AI is writing all the reviews: I bring good news
14
24
3
527
63,276
EAGer mentioned
Cohere Labs
@Cohere_LabsJul 6
We’re thrilled to share that research from Cohere Labs and @Cohere will be presented at ICML this week! 🥳 📄ModernVBERT: Towards Smaller Visual Document Retrievers. Congrats to authors Paul Teiletche Quentin Mace, Max Conti, Antonio Loison, Gautier Viaud, @PierreColombo6, Manuel Faysse (arxiv.org/pdf/2510.01149) 🔎Investigating continual pretraining in large language models: Insights and implications. Congrats to authors Çağatay Yıldız, Nishaanth Kanna Ravichandran, Nitin Sharma, Matthias Bethge, @beyzaermis (openreview.net/pdf?id=aKjJoE…) ⌚EAGER: Entropy-Aware GEneRation for Adaptive Inference-Time Scaling. Congrats to authors @daniel_sc4, Leonidas Zotos, Elisabetta Fersini, and Malvina Nissim, @ahmetustun89 (arxiv.org/abs/2510.11170) 📈When AI Benchmarks Plateau: A Systematic Study of Benchmark Saturation. Congrats to all 35+ authors including Cohere's @YiyangNan (arxiv.org/abs/2602.16763)
1
224
Heading to ICML 🇰🇷 and finding Seoul’s metro map a bit too branchy? Check out my metro poster instead! We branch only when an LM is uncertain, improving both performance and efficiency on reasoning tasks. (applying the same strategy to public transit is not guaranteed to help)
3
5
19
1,237
I’ll be presenting this work at ICML on Wednesday. If you’re curious, the paper is already available: arxiv.org/abs/2510.11170 Happy to chat about reasoning, RL, agents, interpretability, or anything adjacent!

EAGer: Entropy-Aware GEneRation for Adaptive Inference-Time Scaling

With the rise of reasoning language models and test-time scaling methods as a paradigm for improving model performance, substantial computation is often required to generate multiple candidate...

arxiv.org
1
138
I'd never have guessed models commit to their final answer this early, often within the first 20% of reasoning, across math/logic tasks and model families. The rest is mostly hedging that doesn't change their mind. And turns out they encode this internally, we can decode it! 🧵👇
Sara Candussio@SaraCandussioJul 1
Are all the CoT steps necessary? In our latest paper, we find evidence for the existence of a commitment boundary, marking a sharp transition from no/mid guesses to the model final answer across various reasoning tasks and model families. Thread 🧵👇
1
9
1,900
Daniel Scalena retweeted
Cohere
@cohereJun 9
Introducing Cohere's first open-source coding model: North Mini Code Small & efficient, designed for agentic performance and built for community input.

Meet North Mini Code

Cohere's first Apache 2.0 coding model

68
263
101
2,268
600,918
personal update: I'll be starting an internship at @cohere, working on code agents, one of the most interesting things happening in AI right now. feeling really grateful for this one and genuinely excited to see where it goes. here we go! 👀
1
1
86
6,435
Really glad to see this work accepted at ICML 2026🎉 Can’t wait to present this work in Seoul!
Daniel Scalena@daniel_sc416 Oct 2025
You can easily save up to 65% of compute while improving performance on reasoning tasks 🤯 👀 Meet EAGer: We show that monitoring token-level uncertainty lets LLMs allocate compute dynamically - spending MORE on hard problems, LESS on easy ones. 🧵👇
1
6
34
3,528
Presenting this tomorrow at EACL2026, poster session at 9am If you’re around come say hi, happy to chat about the work and ideas More details in the thread 👇
Daniel Scalena@daniel_sc423 May 2025
📢 New paper: Applied interpretability 🤝 MT personalization! We steer LLM generations to mimic human translator styles on literary novels in 7 languages. 📚 SAE steering can beat few-shot prompting, leading to better personalization while maintaining quality. 🧵1/
1
6
1,038
Daniel Scalena retweeted
Gabriele Sarti
@gsarti_Jan 9
Happy to announce I will be mentoring a SPAR project this Spring! ✨Check out the programme and apply by Jan 14th to work with me on understanding and mitigating implicit personalization in LLMs, i.e. how models form hidden beliefs about users that shape their responses.
1
2
1
15
855
Daniel Scalena retweeted
Gabriele Sarti
@gsarti_Jan 4
Now accepted at EACL main! Check it out! ⬇️
Daniel Scalena@daniel_sc423 May 2025
📢 New paper: Applied interpretability 🤝 MT personalization! We steer LLM generations to mimic human translator styles on literary novels in 7 languages. 📚 SAE steering can beat few-shot prompting, leading to better personalization while maintaining quality. 🧵1/
1
15
1,454
Want models to translate in the style you actually like? Our paper just got accepted at EACL Main 🚀, check out our work on using interpretability for MT personalization! And, see you in Morocco! 🇲🇦
Daniel Scalena@daniel_sc423 May 2025
📢 New paper: Applied interpretability 🤝 MT personalization! We steer LLM generations to mimic human translator styles on literary novels in 7 languages. 📚 SAE steering can beat few-shot prompting, leading to better personalization while maintaining quality. 🧵1/
3
281
Takeaway: EAGer shows we can be MORE efficient & MORE effective by letting models focus compute where it matters most. 📄Paper: arxiv.org/abs/2510.11170 💻Code: github.com/DanielSc4/EAGer ✨Huge thanks to my mentors and collaborators @ZotosLeo @FersiniE @MalvinaNissim @ahmetustun89

EAGer: Entropy-Aware GEneRation for Adaptive Inference-Time Scaling

With the rise of reasoning language models and test-time scaling methods as a paradigm for improving model performance, substantial computation is often required to generate multiple candidate...

arxiv.org
1
1
149
Also cc @jiawzhao and @FuYichao123 — I guess your great work on DeepConf is highly relevant to this project!
138
Had so much fun working on this! Thank you @Cohere_Labs for sharing our work!
Cohere Labs
@Cohere_Labs16 Oct 2025
How can we make reasoning models more efficient without sacrificing performance?  Introducing EAGER, our new entropy-aware generation method, saving compute by up to 65% while lifting Pass@k by up to 37% on benchmarks like AIME.
1
6
433
Load more