Aditi Mavalankar
@aditimavalankar
Followers
2K
Following
1K
Media
8
Statuses
133
Research Scientist @DeepMind working on Gemini Thinking
London, UK
Joined March 2017
Excited to share our recent work, AuPair, an inference-time technique that builds on the premise of in-context learning to improve LLM coding performance!.
arxiv.org
Scaling up inference-time compute has proven to be a valuable strategy in improving the performance of Large Language Models (LLMs) without fine-tuning. An important task that can benefit from...
3
4
41
Gemini with advanced deep think achieved gold medal-level performance at IMO 2025!🥇. Very happy to have been a small part of this collaboration on the inference side, and congrats to everyone involved!.
An advanced version of Gemini with Deep Think has officially achieved gold medal-level performance at the International Mathematical Olympiad. 🥇. It solved 5️⃣ out of 6️⃣ exceptionally difficult problems, involving algebra, combinatorics, geometry and number theory. Here’s how 🧵
0
1
19
On my way to #ICML2025 to present our algorithm that strongly scales with inference compute, in both performance and sample diversity! 🚀. Reach out if you’d like to chat more!
Excited to share our recent work, AuPair, an inference-time technique that builds on the premise of in-context learning to improve LLM coding performance!.
2
9
82
Accepted to #ICML2025 .See you in Vancouver!.
Excited to share our recent work, AuPair, an inference-time technique that builds on the premise of in-context learning to improve LLM coding performance!.
0
1
17
This was a really fun collaboration with my brilliant collaborators Hassan Mansoor, Zita Marinho, Masha Samsikova, and Tom Schaul!.
0
0
5
In addition to this, AuPair has been shown to work better across codeforces difficulty levels and preserve coverage of problem categories from the training data distribution (see paper for more details).
1
0
2
4) the responses produced by the model have high diversity for the more performant models.
1
0
2
3) our approach exhibits strong scaling with inference-time compute, and even after 100+ LLM calls, we do not see plateauing in the scaling curve;
1
0
2
2) we observe strong generalisation across datasets and models, implying that the process of curating these examples can be performed once and the benefits in performance can be reaped multiple times;
1
0
3
Injecting different examples into the prompt has several benefits: 1) we see significant gains in performance compared to best-of-N and self-repair baselines on multiple model families: Gemini, Gemma, and GPT;
1
0
5
Fun fact: the title “AuPair” has multiple interpretations: at a higher level, it guides LLMs to better behaviour with a predefined set of examples; it is also a conjunction of Au, the chemical symbol for gold, and pair, i.e. golden pairs!.
1
0
5
For the coding domain, a golden example pair, or AuPair, contains the problem description, an incorrect guess, and a fix that improves the solution.
1
0
5
Our submodular approach yields a fixed ordered set of complementary and useful AuPairs. For a budget of N LLM calls, the model is given N different prompts to answer the same question, where each prompt contains a different golden example.
1
0
6
The key idea underlying our approach is simple: our approach curates a fixed set of golden examples (AuPairs) provided as 1-shot in-context examples during inference. We show that using AuPairs significantly improves code repair performance and scales well with inference compute!.
1
0
4
Outstanding achievement, congratulations, @demishassabis and John Jumper!! 🎉.
Huge congratulations to @DemisHassabis and John Jumper on being awarded the 2024 Nobel Prize in Chemistry for protein structure prediction with #AlphaFold, along with David Baker for computational protein design. This is a monumental achievement for AI, for computational.
0
0
22
RT @demishassabis: Advanced mathematical reasoning is a critical capability for modern AI. Today we announce a major milestone in a longsta….
0
580
0
RT @MichaelD1729: Presenting new work clarifying our perspective on open-endedness (co-lead @edwardfhughes). A clear consequence of our def….
0
5
0
RT @demishassabis: Delighted and honoured to receive a Knighthood for services to AI. It’s been an incredible journey so far building @Goog….
0
290
0
RT @_rockt: I am really excited to reveal what @GoogleDeepMind's Open Endedness Team has been up to 🚀. We introduce Genie 🧞, a foundation….
0
552
0
If this is the way @British_Airways treats business class customers, I wonder what kind of treatment I should expect if I had been flying economy class as I usually do. Threatening foreigners with security when the fault is 100% on the airline is clearly wrong. Do better!.
2
0
29