Harvard Data-Centric Machine Learning Group

university

https://dmelis.github.io

AI & ML interests

Data-Centric ML

Recent Activity

nihalnayak authored a paper 7 days ago

TAGLETS: A System for Automatic Semi-Supervised Learning with Auxiliary Data

nihalnayak authored a paper 7 days ago

PromptSource: An Integrated Development Environment and Repository for Natural Language Prompts

nihalnayak authored a paper 7 days ago

BLOOM: A 176B-Parameter Open-Access Multilingual Language Model

View all activity

Papers

A Critical Look at Targeted Instruction Selection: Disentangling What Matters (and What Doesn't)

Boomerang Distillation Enables Zero-Shot Model Size Interpolation

View all Papers

authored 6 papers 7 days ago

TAGLETS: A System for Automatic Semi-Supervised Learning with Auxiliary Data

Paper • 2111.04798 • Published Nov 8, 2021

PromptSource: An Integrated Development Environment and Repository for Natural Language Prompts

Paper • 2202.01279 • Published Feb 2, 2022

BLOOM: A 176B-Parameter Open-Access Multilingual Language Model

Paper • 2211.05100 • Published Nov 9, 2022 • 40

K-Paths: Reasoning over Graph Paths for Drug Repurposing and Drug Interaction Prediction

Paper • 2502.13344 • Published Feb 18, 2025

$100K or 100 Days: Trade-offs when Pre-Training with Academic Resources

Paper • 2410.23261 • Published Oct 30, 2024

Thinking at the Right Size: Amortized Distillation Across Post-Trained LLMs

Paper • 2608.22854 • Published 26 days ago

updated 2 models 12 days ago

Harvard-DCML/ADAPT-Qwen3-8.5B

Text Generation • 0.5B • Updated 12 days ago • 312

Harvard-DCML/ADAPT-Qwen3-8.5B-Base

Text Generation • 0.5B • Updated 12 days ago • 315

updated a collection 13 days ago

ADAPT

Distilled models from paper Thinking at the Right Size: Amortized Distillation Across Post-Trained LLMs (https://arxiv.org/abs/2608.22854). • 7 items • Updated 13 days ago

published a model 13 days ago

Harvard-DCML/ADAPT-Qwen3-8.5B

Text Generation • 0.5B • Updated 12 days ago • 312

updated a collection 13 days ago

ADAPT

Distilled models from paper Thinking at the Right Size: Amortized Distillation Across Post-Trained LLMs (https://arxiv.org/abs/2608.22854). • 7 items • Updated 13 days ago

published a model 13 days ago

Harvard-DCML/ADAPT-Qwen3-8.5B-Base

Text Generation • 0.5B • Updated 12 days ago • 315

updated a model 13 days ago

Harvard-DCML/ADAPT-Llama3.1-4.8B-Instruct

Text Generation • 1B • Updated 13 days ago • 479

published a model 13 days ago

Harvard-DCML/ADAPT-Llama3.1-4.8B-Instruct

Text Generation • 1B • Updated 13 days ago • 479

updated a model 13 days ago

Harvard-DCML/ADAPT-Olmo3-4.3B-Instruct

Text Generation • 1B • Updated 13 days ago • 466

published a model 13 days ago

Harvard-DCML/ADAPT-Olmo3-4.3B-Instruct

Text Generation • 1B • Updated 13 days ago • 466

updated a model 13 days ago

Harvard-DCML/ADAPT-Qwen3-2.3B-Instruct

Text Generation • 0.6B • Updated 13 days ago • 477