PhD student at @EPFL; prev AITHYRA, CS at KAIST; LLM/continuous/flow reasoning; geometric deep learning
Lausanne, Switzerland
Joined May 2023
- Tweets117
- Following648
- Followers476
- Likes382
Pinned Tweet
Can neural networks learn to think by denoising?
Excited to share Thinking with Looped Flows!
We train recurrent reasoning with local denoising objectives. It is a new approach for reasoning that achieves SoTA on ARC-AGI-1 and 2 among looped models.
In collaboration with @chandavidlee, @FEijkelboom, @nmboffi, @ismaililkanc, @jw9730.
1/🧵
Ayhan Suleymanzade retweeted
This will change how LLMs are trained: Parallel Language Models predict several tokens in parallel at autoregressive quality. Add a parallel decoder at pretraining, get 5.7x speedup after posttraining.
🧐 → arxiv.org/abs/2610.26693
ℹ️ → justuswill.com/plm/
Read on: 🧵 1/4
Ayhan Suleymanzade retweeted
Looking forward to speaking at the BLISS AI Speaker Series next Tuesday! 🧠
I’ll talk about our recent projects at the intersection of AI and neuroscience. If you’re in Berlin, come say hi!
🗓️ Tue, Nov 3, 7pm
📍 TU Berlin, Room C130
🎟️ Free: lnkd.in/dyPshYJB
Ayhan Suleymanzade retweeted
Happy to share our new work: Distributional Functional Rewards (DFR)! 🎉
We improve diversity and fidelity in diffusion fine-tuning 🔧✨
Big thanks to my coauthors @JeongjaeLee0, Jong Chul Ye, @mmbronstein, @Yoshua_Bengio, @AlexanderTong7 and @danyalrehman17! 🙏 Thread below 👇
🚨 New paper alert!
Despite recent advances in reward fine-tuning for diffusion and flow models, achieving high rewards remains challenging without sacrificing sample diversity or substantially drifting away from the pretrained distribution.
We address this with DFR: Distributional Functional Rewards, a framework that balances reward optimization with preserving desirable properties of the generated distribution! 🧵
Ayhan Suleymanzade retweeted
♟♥️⚽️ Chess helps make dreams come true!
Harry Kane shared his love for the game in an interview with GQ.
📷 Harry Kane's social media
Ayhan Suleymanzade retweeted
Make sure to check out @mariagrandury 's Oral presentation about INCLUDE-2.0 today at 11 AM at @emnlpmeeting #NLProc !
Ayhan Suleymanzade retweeted
We're excited to release INCLUDE-2.0! 🌍 An expanded multilingual benchmark evaluating models' regional factual knowledge using questions from local exams across 91 languages, 22 scripts and 86 countries.
Ayhan Suleymanzade retweeted
Super excited to present INCLUDE-2.0 at #EMNLP2026 🎉
Join me tomorrow at 12:15 in room G13 for our oral presentation in the Resources and Evaluation session!
Come to chat about the paper or collaborate in future INCLUDE versions!
Ayhan Suleymanzade retweeted
Rumor has it frontier labs are using looped LMs to scale quality without scaling parameters. But KV-cache memory still grows with every loop. Is that necessary?
No! Our method BeNice! unlocks near-lossless sharing of the final KV-cache across loops: more depth, same KV memory.
This lossless KV-sharing isn’t guaranteed out of the box with looped models. 👀👇for our method BeNice! to teach your model that sharing is caring!
Ayhan Suleymanzade retweeted
Hinton is giving the Gödel lecture at TU Wien and this is our current state according to him:
Ayhan Suleymanzade retweeted
Frontier LLMs can tackle hard problems like Navier-Stokes. But can they solve proofs that are 22 steps deep? 🤔
Meet ProloNg, a Prolog benchmark to evaluate how deep LLMs can go.
Most of the 8 models we test approach chance beyond depth 10 📉
🧵1/7
Ayhan Suleymanzade retweeted
Can a language model be based purely on self-conditioning? 🤔
Yes! Meet LAW 👋 (a.k.a. Loopholing All the Way!)
🧵1/8
Ayhan Suleymanzade retweeted
🚨 New paper!
What if LLMs could reason like we do when solving Sudoku: keep a few candidates per cell and iteratively narrow them down?
Introducing Binary Discrete Diffusion for language and reasoning tasks: it denoises sets of candidate tokens using binary representations! 🧵
Ayhan Suleymanzade retweeted
Excited to announce the follow-up to our NeurIPS 2026 Oral: Improved Training of Strong Stochastic Flow Maps!
We improve the training of SSFMs by unlocking derivative-based losses!
Joint work with @antoniofrancaib @rmincam @ChristianKoke @Niklas_TR and @AlexanderTong7
Happy to see the use of looped flows in science 🧬 :) Great work @DidiKieran!
Ayhan Suleymanzade retweeted
Can flow language models scale to billions of parameters and handle real generative tasks?
Introducing 🌊 Lafa (Large flow language model), a flow LM scaled up to 8B that matches AR quality, with 🚀 2.5× AR throughput at 7B.
Co-led with @ArnavChavan6 @UdbhavBamba.
Joint work with @wognsfjq96 @hongseu33 @AdtRaghunathan @jw9730
Ayhan Suleymanzade retweeted
🚨 New Paper Alert, 🎤 Accepted for an oral presentation at the NeurIPS 2026 STODY workshop!
📣 Introducing: Normalizing Flow Maps (NFM): Combining few-step transport from Flow maps with tractable likelihoods from Normalizing Flows.
This is my first work with my first crop of new PhD students, @francescoshox and @tim_hitge, who genuinely knocked this project out of the park.
📄 Paper: arxiv.org/abs/2610.22711
📝 Blog: the-bose-lab.github.io/norma…
💻 Code: github.com/francescoshox/Nor…
🧵
1/8
Ayhan Suleymanzade retweeted
One model, wildly different graphs and tasks, and some very intriguing generalization capabilities.
Excited about the future of this! Fantastic work by @louistichelman with great contributions from the usual suspects @hxyscott and @jw9730!
Ayhan Suleymanzade retweeted
We’re bringing natural-language control to any scientific generator.
We are very excited to introduce LLM-to-Flow Distillation (LFD), a method to adapt scientific generators to high-level natural-language discovery goals.
Starting from textual goals like “design peptides with promising anticancer activity” or “design small molecules binding to the D2 dopamine receptor”, LFD distills high-level goal-related knowledge from LLMs into scientific generators through iterative DPO on LLM-generated sub-goals.
🔥 LFD does not need task-specific proxy rewards, works with any frontier LLM via API-access, and outperforms state-of-the-art reward-guided methods such as FlowGRPO and DiffusionNFT on held-out computational evaluations. (1/6)
Ayhan Suleymanzade retweeted
Our new preprint Depth Memory is finally live!
We propose a framework that unifies residual methods such as mHC and AttnRes and propose Linear Depth Memory, a novel depth mixer. Check it out!
Interactive blog post: phnazari.github.io/blog/2026…
mHC and AttnRes demonstrate that richer depth mixing improves model performance at scale. Is there a unifying perspective? 🤔
Meet our Depth Memory framework: each layer reads from a set of streams to form its input, then writes back its output.
Linear Depth Memory (LDM) equips models with a matrix-valued memory.