dheeraj rajagopal
@dheerajgopal
Followers
486
Following
1K
Media
1
Statuses
144
Joined April 2010
RT @TechCrunch: Fastino trains AI models on cheap gaming GPUs and just raised $17.5M led by Khosla | TechCrunch
0
23
0
0
7
0
RT @tylerachang: Presenting our work on training data attribution for pretraining this morning: -- come stop by in….
0
5
0
RT @shocheen: Really excited for this paper to be out. This project began nearly a year ago when I was at Ai2. Activation steering and rela….
0
6
0
Really proud to share this work by @patrickqdasilva , providing a thorough evaluation of steering LMs across model sizes.
Steering language models by directly intervening on internal activations is appealing–but does it generalize?. We study 3 popular steering methods with 36 models from 14 families (1.5-70B), exposing brittle performance and fundamental flaws in underlying assumptions.🧵👇.(1/10)
0
0
6
RT @NeelNanda5: Apps are open for my MATS stream, where I try to teach how to do great mech interp research. Due Feb 28!. I love mentoring….
0
30
0
Stoked to share our new work on scaling training data attribution (TDA) toward LLM pretraining - and great insights we found along the way! . and more in the thread below from our excellent student researcher @tylerachang.
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
16
RT @tylerachang: We scaled training data attribution (TDA) methods ~1000x to find influential pretraining examples for thousands of queries….
0
20
0
RT @vidhisha_b: Today’s benchmarks evaluate models on more complex tasks involving many skills. But, aggregate benchmark evaluations really….
0
2
0
RT @vidhisha_b: On some personal news: I joined MSR AI Frontiers a few months ago and am very excited to share my first work with this amaz….
0
11
0
RT @iftenney: Super excited for the Gemma model release, and with it a new debugging tool we built on 🔥LIT - use gradient-based salience to….
0
4
0
RT @vidhisha_b: Our Knowledge Card paper was accepted to #ICLR2024 as an Oral 🎊 Really excited to push further on Modular LLMs!.
0
3
0
RT @ghandeharioun: 🧵Can we “ask” an LLM to “translate” its own hidden representations into natural language? We propose 🩺Patchscopes, a new….
0
147
0
RT @ghandeharioun: Research Opportunities @Google-Please RT.Excited to be at #NeurIPS2023! If interested in how model #interpretability can….
0
26
0
RT @chiragraman: Excited to share our #NeurIPS2023 paper "Why Did This Model Forecast This Future?" Our contrastive explanation framework f….
0
7
0
Really thrilled for Ankush’s new lab at BU! He is an excellent researcher and any student would be so lucky to have him as an advisor.
Thrilled to share that I will start a tenure-track assistant professor position in the CS department at Boston University! I am looking for PhD students in the area of programming languages with applications to distributed systems, cryptography, and machine learning.
0
0
2
RT @Das8Ankush: Thrilled to share that I will start a tenure-track assistant professor position in the CS department at Boston University!….
0
29
0
RT @dongyeopkang: I thought LLM ignores diverse opinions bcs it is biased towards aggregating probabilities of the majority. But, we found….
0
2
0