Based in Rome, GLADIA is a team of computer scientists, physicists, engineers and mathematicians venturing beyond the boundaries of machine intelligence

Rome
Joined May 2025
Introducing the GLADIA Research Lab.
6
7
5
100
52,997
GLADIA Research Lab retweeted
worried that somebody might steal your millenium problem proof? You should. In our new paper, HEAT, we make a concrete step in making LLM's private inference stable and practical. (1/6)
1
3
3
124
Original or AI reconstruction? 🎧 You’ll see every switch. Can you hear one? Meet SAGE, our 105M music autoencoder: ⚡ As fast as Stable Audio Open’s VAE 🎧 On par with SAME-L in listening tests, at 1/8 the size 🏆 SOTA on 23/25 fidelity scores + 19/19 semantic tasks (1/6)
1
7
22
997
Reconstruction on 5 benchmarks: • lowest FAD / highest CLAP similarity on 23 of 25 measurements • MUSHRA 81.6, statistically identical to SAME-L (81.8, 852M params, 4× slower) • +17 points over Stable Audio Open at the same speed Trained only on public music. (5/6)
1
2
71
GLADIA Research Lab retweeted
Flywheel is going local. Introducing Flywheel Repos. You can now work on Flywheel graphs directly on your machine, whether you’re online or offline. Make changes locally, and asynchronously push when you’re ready to save or publish them to your online Flywheel graphs. Scientific work is iterative. We've designed Flywheel Repos to give that scientific work room to develop privately. More in this direction soon.
6
6
3
95
9,621
GLADIA Research Lab retweeted
Introducing Limite 1B - Violetto. A model for high-frequency mathematical intelligence.
66
232
80
1,868
296,779
GLADIA Research Lab retweeted
Can a sentence carry a sound? In Communicating Sound Through Natural Language, we introduce lexical acoustic coding (LAC): a way for LLM agents to transmit short sounds as structured English, then re-render the same audio back from that text. (1/6)
5
26
5
147
16,861
GLADIA Research Lab retweeted
research labs should have more physical artifacts. last year, my PhD student @DMarincione carved our @GladiaLab logo into a seal stamp in Korea. we're bringing it back to @icmlconf 2026. let's meet to get your badge stamped! forehead stamping evaluated case by case.
1
4
18
896
GLADIA Research Lab retweeted
Time’s up! Congratulations to @prasannakotyal, @SarahLacard and @amit05prakash, we’ll be reaching out in DMs for coordinating your rewards
1:15 hrs left to the deadline if anyone wants to try sniping the leaderboard 👀
6
3
4
29
5,181
GLADIA Research Lab retweeted
this function is spinning at 154 RPM. that's the tempo of the music; learned with zero rhythm supervision. meet PHALAR, our new audio representation framework. Thread below 🧵 #ICML2026 #MIR 1/6
1
12
2
32
3,505
GLADIA Research Lab retweeted
Today we’re announcing a new campaign in collaboration with @myainotez, with a 1000 USD prize: KernelGuard
3
4
6
55
18,564
Meet us in room 212 starting at 3:30 pm, and bring with you your perspective on the future of research. Our panelists: @FrancescoLocat8 @GabriCorso @FazlBarez  @besanushi @TuhinChakr
1
4
614
What will the role of researchers be in 5 years? What happens to narrow scientific foundation models? Can we really scale our way to genius creativity? We will attempt to answer these questions (and more) on Sunday in the post-agi workshop at @iclr_conf
1
8
3
26
11,290
GLADIA Research Lab retweeted
alright we can't hide it anymore come and see the pringle in all its glory at @iclr_conf
LLMs are injective and invertible. In our new paper, we show that different prompts always map to different embeddings, and this property can be used to recover input tokens from individual embeddings in latent space. (1/6)
3
76
4
1,125
90,687
GLADIA Research Lab retweeted
formally proved in Lean4 that music is not Turing complete. that's because there can be no endogenous self-replication during music playback (unfortunately). proof: every infinite playback of any valid midi-like symbolic composition is eventually periodic.
48
67
8
752
69,811