Tolga Bolukbasi
@tolgab0
Followers
302
Following
417
Media
3
Statuses
92
AI research/Gemini pretraining @GoogleDeepmind, PhD, opinions my own.
Joined November 2014
Our work on scaling training data attribution is out. There are a lot of insights in there, I especially like the distinction between attribution and influence. Thanks to our amazing student researcher Tyler for making this happen.
We scaled training data attribution (TDA) methods ~1000x to find influential pretraining examples for thousands of queries in an 8B-parameter LLM over the entire 160B-token C4 corpus!.
0
1
11
RT @sundarpichai: Our latest Gemini 2.5 Pro update is now in preview. It’s better at coding, reasoning, science + math, shows improved per….
0
751
0
RT @tylerachang: Presenting our work on training data attribution for pretraining this morning: -- come stop by in….
0
5
0
RT @NoamShazeer: This model’s “thinking” capabilities are driving major gains:. 🧑🔬Top performance on math and science benchmarks (AIME, GP….
0
18
0
RT @suchenzang: "From Figure 3(a), it is apparent that many of the benchmarks we considered are substantially cont….
0
16
0
RT @morteymike: @Nexuist I worked on the M series while at Apple. The main advantage that stuck out to me was actually that they were able….
0
486
0
RT @andrew_ilyas: Machine unlearning ("removing" training data from a trained ML model) is a hard, important problem. Datamodel Matching (….
0
23
0
I will be in ATTRIB workshop tomorrow (. Stop by if you’d like to chat with me and connect with other great researchers in this area.
0
1
9
I have been thinking about this since ChatGPT came out. Using RLHF never fully made sense to me given how restricted it is compared to regular RL. There should be a way simpler non-exploring method to distill RM knowledge into the main model.
# RLHF is just barely RL. Reinforcement Learning from Human Feedback (RLHF) is the third (and last) major stage of training an LLM, after pretraining and supervised finetuning (SFT). My rant on RLHF is that it is just barely RL, in a way that I think is not too widely
0
0
1
RT @melvinjohnsonp: Great to see Gemini 1.5 doing well on this new video understanding benchmark!.
0
4
0
It was great to work with Minsuk and excited to see this released. Looking at individual model outputs this way helps one see which examples/tasks are truly wins across model versions and which ones are just due to randomness of generation or raters.
Very excited to open-source LLM Comparator!.This new #visualization tool lets you analyze LLM responses side-by-side. It’s been used for evaluating LLMs @Google, and we're proud to release it as part of Google's Responsible GenAI Toolkit.
0
0
7
RT @kelvin_guu: Great new work from our team and colleagues at.@GoogleDeepMind! On the Massive Text Embedding Benchmark (MTEB), Gecko is th….
0
6
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