WIth Lean-FIRE, we achieved the first end-to-end autoformalization for 13 Putnam problems, but also showed that conjecturing is still a massive reasoning challenge.

We're tackling this by treating conjecturing as its own task. We introduce ConjectureBench to measure performance, and Lean-FIRE, a new method to guide LLMs to generate better conjectures.

New preprint on Formal Math Autoformalization \w Jasivan Sivakumar, @pnborchert, @ronaldca2: https://t.co/bHtgZs5zWD Formal math reasoning of today's LLMs is weaker than we think. Benchmarks often give the solution upfront. In reality, conjecturing is a critical reasoning step!

We are exhibiting at #ACL2025 in Vienna from July 28th to 30th!! Come visit us at booth 14 and have a chat with many of our experts (and me!), especially if interested in our hiring opportunities!! https://t.co/fcbq95Pwn1

1) Delighted to introduce our latest work🥳 (under review) 🙃 🔗 https://t.co/lX7rIQLdk0 🔗 We propose **DReSD: Dense Retrieval for Speculative Decoding** with @HuiyinXue and @glampouras_NLP

"Code-Optimise: Self-Generated Preference Data for Correctness and Efficiency" with Leonidas Gee, @milangritta, Ignacio Iacobacci NAACL 2025 (Findings) [Paper]

"Mixture of Attentions for Speculative Decoding" with @matthieu_zimmer, @milangritta, @hbouammar, Jun Wang ICLR 2025 [Paper] https://t.co/5N2XT6Pyeu [Code]

"Human-like Episodic Memory for Infinite Context LLMs" with @zfountas, Martin Benfeghoul, Adnan Oomerjee, @fenchri, @hbouammar, Jun Wang ICLR 2025 [Paper] https://t.co/khvJfgGAzI [Code]

A strong start of the year with some great paper acceptances, congratulations to all my collaborators! For our less fortunate submissions: improve, rewrite, and let's try again :) #ICLR2025 #NAACL2025 Specifics below :D

Also Internship Positions for the London NLP Team are open year-round. Apply below: https://t.co/ifG7G1ZdGD

Our London NLP team is exploring some exciting and ambitious research directions on LLM efficiency, inference and reasoning. Join us and help us realize them! Take a look at our Research Engineer listing for more details :) https://t.co/9yzVakw82f

Better late than ever, finally, code is available🙂 💻 https://t.co/bsgZsbq5Oh

I am excited to present Agent K as the first end-to-end agent (i.e., autonomous from Kaggle URL to submissions that win competitions) to achieve an equivalent of Kaggle grandmaster level. Our agent codes the whole data science pipeline from a natural language description of the

Research Scientist (permanent) positions just opened in our NLP team in London - Huawei Noah's Ark Lab! Looking for experienced researchers to help us tackle some interesting (and persistent) questions :) Take a look and apply if interested: https://t.co/Q4JajEMFLQ

NAACL was a while ago, but we finally have the code and data publicly available for this!! Check it out!

Some insights from our latest pre-print!!

Excited to present MuLAn at #CVPR2024! Our dataset addresses compositional text2image generation with multi-layer, instance-wise RGBA decomposition annotations for richer multi-modal generation & editing. Poster #275 - FRI 10:30 am https://t.co/qb6N1EDpOv with @DTudosiu et al.

So happy to be back to MCR for this talk! Thanks again for the invitation!

Full-time and part-time internship positions are also open year-round, and we are flexible on starting dates and duration (within reason of' course :). https://t.co/ifG7G1ZdGD

Looking to grow our NLP team in London - Huawei Noah's Ark Lab! We have positions for Research Scientists (permanents) and Engineers (contractors), where they will conduct academic and applied research in NLP and ML. Details below: https://t.co/dDxE97BgRl https://t.co/LGN9snl3yb