3 citations · 11 across the 22 of their papers we have counts for
24 papers
Off-Policy Merging Beats On-Policy Self-Distillation for Continual Learning
Chen Henry Wu, Thomas Zhang, Aditi Raghunathan
A long-standing goal of AI is a model that can continually learn and improve itself. On post-trained models, supervised finetuning (SFT) on new data often causes poor generalizatio…
Natively Unlearnable Large Language Models
Gaurav R. Ghosal, Pratyush Maini, Aditi Raghunathan
Unlearning aims to remove the influence of specific training data sources, but this has proved challenging because the contributions of different sources are entangled within the m…
Annotations Mitigate Post-Training Mode Collapse
Jacob Mitchell Springer, Madhu Advani, Lukas Aichberger +7
Post-training (via supervised fine-tuning) improves instruction-following, but often induces semantic mode collapse by biasing models toward low-entropy fine-tuning data at the exp…
Sharpness-Aware Pretraining Mitigates Catastrophic Forgetting
Ishaan Watts, Catherine Li, Sachin Goyal +2
Pretraining optimizers are tuned to produce the strongest possible base model, on the assumption that a stronger starting point yields a stronger model after subsequent changes lik…
Early Data Exposure Improves Robustness to Subsequent Fine-Tuning
Lawrence Feng, Gaurav R. Ghosal, Jacob Mitchell Springer +2
How can we train models whose post-trained capabilities survive subsequent fine-tuning? Rather than focusing on downstream interventions to mitigate forgetting of upstream capabili…
The Finetuner's Fallacy: When to Pretrain with Your Finetuning Data
Christina Baek, Ricardo Pio Monti, David Schwab +31
Real-world model deployments demand strong performance on narrow domains where data is often scarce. Typically, practitioners finetune models to specialize them, but this risks ove…
Mode-Conditioning Unlocks Superior Test-Time Scaling
Chen Henry Wu, Sachin Goyal, Aditi Raghunathan
Parallel sampling promises substantial gains in test-time scaling, but its effectiveness is sharply limited by diversity collapse, where models concentrate on a few modes and repea…
Mitigating Modal Imbalance in Multimodal Reasoning
Chen Henry Wu, Neil Kale, Aditi Raghunathan
Foundation models (FMs) deployed in real-world tasks such as computer-use agents must integrate diverse modalities. How good are FMs at performing joint reasoning, simultaneously r…
Understanding the Influence of Synthetic Data for Text Embedders
Jacob Mitchell Springer, Vaibhav Adlakha, Siva Reddy +2
Recent progress in developing general purpose text embedders has been driven by training on ever-growing corpora of synthetic LLM-generated data. Nonetheless, no publicly available…
Memorization Sinks: Isolating Memorization during LLM Training
Gaurav R. Ghosal, Pratyush Maini, Aditi Raghunathan
Large language models are susceptible to memorizing repeated sequences, posing privacy and copyright concerns. A popular mitigation strategy is to remove memorized information from…
Reasoning as an Adaptive Defense for Safety
Taeyoun Kim, Fahim Tajwar, Aditi Raghunathan +1
Reasoning methods that adaptively allocate test-time compute have advanced LLM performance on easy to verify domains such as math and code. In this work, we study how to utilize th…
Exact Unlearning of Finetuning Data via Model Merging at Scale
Kevin Kuo, Amrith Setlur, Kartik Srinivas +2
Approximate unlearning has gained popularity as an approach to efficiently update an LLM so that it behaves (roughly) as if it was not trained on a subset of data to begin with. Ho…
Roll the dice & look before you leap: Going beyond the creative limits of next-token prediction
Vaishnavh Nagarajan, Chen Henry Wu, Charles Ding +1
We design a suite of minimal algorithmic tasks that are a loose abstraction of open-ended real-world tasks. This allows us to cleanly and controllably quantify the creative limits…
Overtrained Language Models Are Harder to Fine-Tune
Jacob Mitchell Springer, Sachin Goyal, Kaiyue Wen +5
Large language models are pre-trained on ever-growing token budgets under the assumption that better pre-training performance translates to improved downstream models. In this work…
Not-Just-Scaling Laws: Towards a Better Understanding of the Downstream Impact of Language Model Design Decisions
Emmy Liu, Amanda Bertsch, Lintang Sutawika +9
Improvements in language model capabilities are often attributed to increasing model size or training data, but in some cases smaller models trained on curated data or with differe…
Mitigating Bias in RAG: Controlling the Embedder
Taeyoun Kim, Jacob Springer, Aditi Raghunathan +1
In retrieval augmented generation (RAG) systems, each individual component -- the LLM, embedder, and corpus -- could introduce biases in the form of skews towards outputting certai…
Understanding Finetuning for Factual Knowledge Extraction
Gaurav Ghosal, Tatsunori Hashimoto, Aditi Raghunathan
In this work, we study the impact of QA fine-tuning data on downstream factuality. We show that fine-tuning on lesser-known facts that are poorly stored during pretraining yields s…
Dissecting Adversarial Robustness of Multimodal LM Agents
Chen Henry Wu, Rishi Shah, Jing Yu Koh +3
As language models (LMs) are used to build autonomous agents in real environments, ensuring their adversarial robustness becomes a critical challenge. Unlike chatbots, agents are c…
Sharpness-Aware Minimization Enhances Feature Quality via Balanced Learning
Jacob Mitchell Springer, Vaishnavh Nagarajan, Aditi Raghunathan
Sharpness-Aware Minimization (SAM) has emerged as a promising alternative optimizer to stochastic gradient descent (SGD). The originally-proposed motivation behind SAM was to bias…
Scaling Laws for Data Filtering -- Data Curation cannot be Compute Agnostic
Sachin Goyal, Pratyush Maini, Zachary C. Lipton +2
Vision-language models (VLMs) are trained for thousands of GPU hours on carefully curated web datasets. In recent times, data curation has gained prominence with several works deve…
Testing the Limits of Jailbreaking Defenses with the Purple Problem
Taeyoun Kim, Suhas Kotha, Aditi Raghunathan
The rise of "jailbreak" attacks on language models has led to a flurry of defenses aimed at preventing undesirable responses. We critically examine the two stages of the defense pi…
Repetition Improves Language Model Embeddings
Jacob Mitchell Springer, Suhas Kotha, Daniel Fried +2
Bidirectional models are considered essential for strong text embeddings. Recent approaches to adapt autoregressive language models (LMs) into strong text embedding models have lar…
Complementary Benefits of Contrastive Learning and Self-Training Under Distribution Shift
Saurabh Garg, Amrith Setlur, Zachary Chase Lipton +3
Self-training and contrastive learning have emerged as leading techniques for incorporating unlabeled data, both under distribution shift (unsupervised domain adaptation) and when…
Multitask Learning Can Improve Worst-Group Outcomes
Atharva Kulkarni, Lucio Dery, Amrith Setlur +3
In order to create machine learning systems that serve a variety of users well, it is vital to not only achieve high average performance but also ensure equitable outcomes across d…