Carlos Muñoz Ferrandis (@Carlos_MFerr) header
Carlos Muñoz Ferrandis (@Carlos_MFerr) profile picture

Alinia AI co-founder I PhD l Views my own

Everywhere
Joined March 2021
Photo by Carlos Muñoz Ferrandis (@Carlos_MFerr)Photo by Carlos Muñoz Ferrandis (@Carlos_MFerr)Photo by Carlos Muñoz Ferrandis (@Carlos_MFerr)Photo by Carlos Muñoz Ferrandis (@Carlos_MFerr)Photo by Carlos Muñoz Ferrandis (@Carlos_MFerr)Photo by Carlos Muñoz Ferrandis (@Carlos_MFerr)Photo by Carlos Muñoz Ferrandis (@Carlos_MFerr)Photo by Carlos Muñoz Ferrandis (@Carlos_MFerr)Photo by Carlos Muñoz Ferrandis (@Carlos_MFerr)Photo by Carlos Muñoz Ferrandis (@Carlos_MFerr)
Heading to San Francisco for @PlugandPlayTC Silicon Valley Summit!
1
1
1
43
Increasing adoption of AI Agents renders legacy governance frameworks, such as the 3 lines of defence system in banking, obsolete. To tackle this systemic challenge we're building adaptive regulatory control intelligence.
13
Carlos Muñoz Ferrandis retweeted
We open sourced Sev huggingface.co/alinia/Sev Matching Jev's ranking took a week. Making its decisions trustworthy is a different problem, one that still takes expert-labeled data and a judge trained on it. alinia.ai/resources/how-far-…
1
8
20
706
We are open sourcing Sev, Alinia AI new guardrail model available on @huggingface , in our mission to push the frontier of AI safety in critical scenarios. Link to Sev repo: huggingface.co/alinia/Sev
1
3
7
151
It took us one week. Two A100s. Zero Alinia data. Only open datasets and synthetic replies. - Beating Jev on 6 of 8 public guard benchmarks. - 0.936 mean AUROC vs Jev’s 0.934 - 90 ms per request on one A100 .
1
49
Consistency is where both Sev and Jev break. Reword the same legal-advice policy and Sev blocks anywhere from 27% to 93% of compliant replies. Jev: 11% to 79%. Open data gets you a Jev, not a reliable judge.
26
Carlos Muñoz Ferrandis retweeted
BAIA, BAIA... un Base Model español que le moja la oreja a modelos frontera en el vertical de regulación e impuestos. Me encantaría ver esto en la #trg26... ¡quién sabe!
And now you have access to Alinia’s legal and tax advice data sets. alinia.ai/resources/benchmar… Our customers also have access to these low-latency policy judges with frontier-level model accuracy
3
7
1
57
7,988
Carlos Muñoz Ferrandis retweeted
#Startup vibes with @Carlos_MFerr 💪🔥
2
1
21
1,736
Carlos Muñoz Ferrandis retweeted
We are now inviting participation from the community to help shape the future of Responsible AI Licenses (RAIL)! RAIL has recently been adopted by many new AI materials (e.g. @BigCodeProject)...
1
8
1
12
9,022
Carlos Muñoz Ferrandis retweeted
It's been an exciting year for 🤗Transformers. We tripled the number of weekly active users over 2022, with over 1M users most weeks now and 300k daily pip installs on average🤯
10
82
5
610
153,092
Carlos Muñoz Ferrandis retweeted
Finally, we summarized our findings in a technical report with a wonderful group of collaborators: Paper: hf.co/datasets/bigcode/admin… So what's next? Scaling to larger models and training more languages early next year! 🚀
1
1
1
23
3,096
Carlos Muñoz Ferrandis retweeted
Announcing a holiday gift: 🎅SantaCoder - a 1.1B multilingual LM for code that outperforms much larger open-source models on both left-to-right generation and infilling! Demo: hf.co/spaces/bigcode/santa-d… Paper: hf.co/datasets/bigcode/admin… Attribution: hf.co/spaces/bigcode/santaco… A🧵:
9
196
28
823
264,129
Carlos Muñoz Ferrandis retweeted
We crossed 100,000 public AI models on the @huggingface hub available for free to all. Thank you to the whole community of contributors. Proud to make ML more open & collaborative!
8
59
2
450
58,952