Severi Rissanen

[Arno Solin @arnosolin]

1/ BayesVLM turns pre-trained VLMs into probabilistic models post hoc, without retraining or fine-tuning. It is model-agnostic, works with standard #VLM architectures, and provides uncertainty estimates with virtually no inference overhead.

[Arno Solin @arnosolin]

1/ 🔥 New paper: Differentiable Vector Quantization (DiVeQ) 🔥

Vector quantization (VQ) is a key building block in modern AI. It links continuous data like images and audio to discrete representations (tokens) used by transformers.

[Prakhar Verma @prakhar_verma7]

Did #NeurIPS2025 reviews pull you down and #ICLR2026 deadline looks too tight?

🚀 Submit to AISTATS 2026.

Abstract deadline: 25 September
Submission deadline: 2 October

Details: https://t.co/qBy3FQac2G

[Prakhar Verma @prakhar_verma7]

The #AISTATS2026 Call for Papers is out!

Abstract Deadline: Sep 25, 2025 (AoE)
Full paper Deadline: Oct 2, 2025 (AoE)
Supplementary Deadline: Oct 9, 2025 (AoE)

It is a privilege to serve as a Workflow Chair alongside @_Mr_YJ and Program Chairs Aaditya Ramdas and @arnosolin

[Tony RuiKang OuYang @TonyRKOuYang]

Exited to share our new paper accepted by ICML 2025
👉 “PTSD: Progressive Tempering Sampler with Diffusion”
, which aims to make sampling from unnormalised densities more efficient than state-of-the-art methods like parallel tempering.

Check our threads below 👇

[Severi Rissanen @SeveriRissanen]

@prlz77 Really interested in this, applied :) I am a fourth-year PhD student, recently working on applying inference-time controls to diffusion models.

[Severi Rissanen @SeveriRissanen]

@CSProfKGD @sedielem @ArashVahdat Thanks for the shout-out :) am quite proud of that figure 😁