Ivan Titov

@iatitov

Professor of Natural Language Processing and ML at Uni Edinburgh / Uni Amsterdam

Edinburgh, Scotland

ivan-titov.org

Joined September 2016

788 Following

7,235 Followers

Replies Reposts Media

Ivan Titov
@iatitov
Nov 2
Back from @emnlpmeeting in Budapest a couple of days ago! Nice to catch up with old friends and meet some new people. Also great to have quite a few people from our group there, we even managed to get some of us together for a photo! @EdinburghNLP @AmsterdamNLP
Ivan Titov
@iatitov
Oct 2
Very happy to see @j0hngou’s work accepted as a NeurIPS oral! I particularly like the broader idea: let models interact richly during training while pressuring them to reduce that interaction. This gives a denser learning signal and leads to more effective one-shot collaboration
John Gkountouras
@j0hngou
Oct 2
Happy to share that our paper "Clarification as Supervision" was accepted as an Oral at #NeurIPS2026 (top ~0.3%). When a text-only reasoner answers from a VLM's caption and gets it wrong, we usually blame the reasoning. Often the caption left out the details the reasoner needed🧵
Ivan Titov
@iatitov
Aug 11
📢 We're hiring! Postdocs at @EdinburghUni to develop novel architectures & learning objectives for open-source foundation models in the SOFAIR Lab - with @frank_e_keller, @mlapata, @AmosStorkey, me + partners at UCL, Cambridge & Oxford. Apply 👇
elxw.fa.em3.oraclecloud.com
Research Associate
Five positions are part of the Science of Fundamental AI Research (SOFAIR) Lab, a partnership between UCL and the Universities of Cambridge, Edinburgh, and Oxford. SOFAIR is one of two new national...
Ivan Titov
@iatitov
Jul 21
Consensus is usually compressed into a vote or scalar reward. This work instead uses a consensus solution as context for a frozen self-teacher, turning agreement into dense token-level supervision. Recovers much of the benefit of gold solutions, without labels
John Gkountouras
@j0hngou
Jul 21
Sampling many solutions and taking the majority answer is the most reliable label-free way to improve LLM reasoning. Training methods reduce the consensus to a filter, a preference, or a scalar reward. What if the model could read them instead?
1
Ivan Titov
@iatitov
Jul 14
On-policy self-distillation: the teacher saw the solution; the student didn't. Copy that confidence and you gain in-domain but lose out-of-domain reasoning. GeoSD scales the teacher's pull by overlap and bounds drift - OOD +5.7–8.6 pts, ID largely intact. Led by @chatruncata
Josip Jukic
@chatruncata
Jul 14
Self-distillation gives LLMs dense reasoning feedback without requiring a larger teacher. But ID gains can come at the cost of OOD generalization. New joint work with @iatitov : "Geometric Self-Distillation for Reasoning Generalization" (arxiv.org/abs/2607.06855)
1

Ivan Titov

Ivan Titov’s posts