skip to main content


Search for: All records

Creators/Authors contains: "Chen, A."

Note: When clicking on a Digital Object Identifier (DOI) number, you will be taken to an external site maintained by the publisher. Some full text articles may not yet be available without a charge during the embargo (administrative interval).
What is a DOI Number?

Some links on this page may take you to non-federal websites. Their policies may differ from this site.

  1. Does language help make sense of the visual world? How important is it to actually see the world rather than having it described with words? These basic questions about the na- ture of intelligence have been difficult to answer because we only had one example of an intelligent system – humans – and limited access to cases that isolated language or vision. How- ever, the development of sophisticated Vision-Language Mod- els (VLMs) by artificial intelligence researchers offers us new opportunities to explore the contributions that language and vi- sion make to learning about the world. We ablate components from the cognitive architecture of these models to identify their contributions to learning new tasks from limited data. We find that a language model leveraging all components recovers a majority of a VLM’s performance, despite its lack of visual in- put, and that language seems to allow this by providing access to prior knowledge and reasoning. 
    more » « less
    Free, publicly-accessible full text available July 24, 2025
  2. The potential for pre-trained large language models (LLMs) to use natural language feedback at inference time has been an exciting recent development. We build upon this observation by formalizing an algorithm for learning from natural language feedback at training time instead, which we call Imitation learning from Language Feedback (ILF). ILF requires only a small amount of human-written feedback during training and does not require the same feedback at test time, making it both user-friendly and sample-efficient. We further show that ILF can be seen as a form of minimizing the KL divergence to the target distribution and demonstrate proof-of-concepts on text summarization and program synthesis tasks. For code generation, ILF improves a Codegen-Mono 6.1B model’s pass@1 rate from 22% to 36% on the MBPP benchmark, outperforming both fine-tuning on MBPP and on human- written repaired programs. For summarization, we show that ILF can be combined with learning from human preferences to improve a GPT-3 model’s summarization performance to be comparable to human quality, outperforming fine-tuning on human-written summaries. Overall, our results suggest that ILF is both more effective and sample-efficient than training exclusively on demonstrations for improving an LLM’s performance on a variety of tasks. 
    more » « less
    Free, publicly-accessible full text available February 28, 2025
  3. Large language models (LLMs) have achieved widespread success on a variety of in-context few shot tasks, but this success is typically evaluated via correctness rather than consistency. We argue that self-consistency is an important criteria for valid multi-step reasoning in tasks where the solution is composed of the answers to multiple sub-steps. We propose two types of self consistency that are particularly important for multi-step reasoning – hypothetical consistency (a model’s ability to predict what its output would be in a hypothetical other context) and compositional consistency (consistency of a model’s final outputs when intermediate sub-steps are replaced with the model’s outputs for those steps). We demonstrate that multiple variants of the GPT-3/-4 models exhibit poor consistency rates across both types of consistency on a variety of tasks. 
    more » « less
    Free, publicly-accessible full text available January 1, 2025
  4. Migrating birds often fly in group formations during the daytime; whereas at night, it is generally presumed that they fly singly. However, it is difficult to quantify group behavior during nocturnal migration as there are few means of directly observing interactions among individuals. We employed an automated form of moonwatching to estimate percentages of birds that appear to migrate in groups during the night within the Central Flyway of North America. We compared percentages of birds in groups across the spring and fall and examined overnight temporal patterns of group behavior. We found groups were rare in both seasons, never exceeding 10% of birds observed, and were almost nonexistent during the fall. We also observed an overnight pattern of group behavior in the spring wherein groups were more commonly detected early in the night and again just before migration activity ceased. This finding may be related to changes in species composition of migrants throughout the night, or alternatively it suggests that group formation may be associated with flocking activity on the ground as groups are most prevalent when birds begin and end a night of migration. 
    more » « less
    Free, publicly-accessible full text available December 15, 2024
  5. We probe the conduction-band offsets (CBOs) and confined states at GaAs/GaAsNBi quantum wells (QWs). Using a combination of capacitance–voltage (C–V) measurements and self-consistent Schrödinger–Poisson simulations based on the effective mass approximation, we identify an N-fraction dependent increase in CBO, consistent with trends predicted by the band anti-crossing model. Using the computed confined electron states in conjunction with photoluminescence spectroscopy data, we show that N mainly influences the conduction band and confined electron states, with a relatively small effect on the valence band and confined hole states in the quaternary QWs. This work provides important insight toward tailoring CBO and confined electron energies, both needed for optimizing infrared optoelectronic devices.

     
    more » « less
  6. Language models (LMs) are pretrained to imitate internet text, including content that would violate human preferences if generated by an LM: falsehoods, offensive comments, personally identifiable information, low-quality or buggy code, and more. Here, we explore alternative objectives for pretraining LMs in a way that also guides them to generate text aligned with human preferences. We benchmark five objectives for pretraining with human feedback across three tasks and study how they affect the trade-off between alignment and capabilities of pretrained LMs. We find a Pareto optimal and simple approach among those we explored: conditional training, or learning distribution over tokens conditional on their human preference scores given by a reward model. Conditional training reduces the rate of undesirable content by up to an order of magnitude, both when generating without a prompt and with an adversarially chosen prompt. Moreover, conditional training maintains the downstream task performance of standard LM pretraining, both before and after task-specific finetuning. Pretraining with human feedback results in much better preference satisfaction than standard LM pretraining followed by finetuning with feedback, i.e., learning and then unlearning undesirable behavior. Our results suggest that we should move beyond imitation learning when pretraining LMs and incorporate human preferences from the start of training. 
    more » « less
  7. Alarcon, Emilio I. (Ed.)
    Membrane proteins (MPs) are essential to many organisms’ major functions. They are notorious for being difficult to isolate and study, and mimicking native conditions for studies in vitro has proved to be a challenge. Lipid nanodiscs are among the most promising platforms for MP reconstitution, but they contain a relatively labile lipid bilayer and their use requires previous protein solubilization in detergent. These limitations have led to the testing of copolymers in new types of nanodisc platforms. Polymer-encased nanodiscs and polymer nanodiscs support functional MPs and address some of the limitations present in other MP reconstitution platforms. In this review, we provide a summary of recent developments in the use of polymers in nanodiscs. 
    more » « less
  8. null (Ed.)