887 citations · 928 across the 20 of their papers we have counts for
26 papers
Position: It's Time to Optimize LLMs for Self-Consistency
Itamar Pres, Belinda Z. Li, Laura Ruis +6
Despite ever-increasing sophistication in language model (LM) pre- and post-training pipelines, many important failures persist: models overcondition on user framing ("sycophancy")…
Introspective Coupling: Self-Explanation Training Tracks Behavioral Change Despite Fixed Supervision
Zifan Carl Guo, Laura Ruis, Jacob Andreas +1
When does training language models (LMs) to generate explanations of their predictions yield faithful introspection, rather than superficial imitation? We study LMs trained to expl…
Explaining Attention with Program Synthesis
Amiri Hayes, Belinda Z Li, Jacob Andreas
A longstanding goal of research on interpretable deep learning is to replace opaque neural computations with human-meaningful symbolic descriptions. In this paper, we propose an ap…
Self-CTRL: Self-Consistency Training with Reinforcement Learning
Itamar Pres, Laura Ruis, Melat Ghebreselassie +2
Language models (LMs) that faithfully describe their own behavior can more easily be audited, understood, and trusted by users. This paper describes Self-Consistency Training with…
Training Language Models to Explain Their Own Computations
Belinda Z. Li, Zifan Carl Guo, Vincent Huang +2
Can language models (LMs) learn to faithfully describe their internal computations? Are they better able to describe themselves than other models? We study the extent to which LMs'…
QuestBench: Can LLMs ask the right question to acquire information in reasoning tasks?
Belinda Z. Li, Been Kim, Zi Wang
Large language models (LLMs) have shown impressive performance on reasoning benchmarks like math and logic. While many works have largely assumed well-defined tasks, real-world que…
(How) Do Language Models Track State?
Belinda Z. Li, Zifan Carl Guo, Jacob Andreas
Transformer language models (LMs) exhibit behaviors -- from storytelling to code generation -- that seem to require tracking the unobserved state of an evolving world. How do they…
Adaptive Language-Guided Abstraction from Contrastive Explanations
Andi Peng, Belinda Z. Li, Ilia Sucholutsky +4
Many approaches to robot learning begin by inferring a reward function from a set of human demonstrations. To learn a good reward, it is necessary to determine which features of th…
Language Modeling with Editable External Knowledge
Belinda Z. Li, Emmy Liu, Alexis Ross +3
When the world changes, so does the text that humans write about it. How do we build language models that can be easily updated to reflect these changes? One popular approach is re…
Bayesian Preference Elicitation with Language Models
Kunal Handa, Yarin Gal, Ellie Pavlick +4
Aligning AI systems to users' interests requires understanding and incorporating humans' complex values and preferences. Recently, language models (LMs) have been used to gather in…
Learning with Language-Guided State Abstractions
Andi Peng, Ilia Sucholutsky, Belinda Z. Li +4
We describe a framework for using natural language to design state abstractions for imitation learning. Generalizable policy learning in high-dimensional observation spaces is faci…
Preference-Conditioned Language-Guided Abstraction
Andi Peng, Andreea Bobu, Belinda Z. Li +5
Learning from demonstrations is a common way for users to teach robots, but it is prone to spurious feature correlations. Recent work constructs state abstractions, i.e. visual rep…
Eliciting Human Preferences with Language Models
Belinda Z. Li, Alex Tamkin, Noah Goodman +1
Language models (LMs) can be directed to perform target tasks by using labeled examples or natural language prompts. But selecting examples or writing prompts for can be challengin…
Toward Interactive Dictation
Belinda Z. Li, Jason Eisner, Adam Pauls +1
Voice dictation is an increasingly important text input modality. Existing systems that allow both dictation and editing-by-voice restrict their command language to flat templates…
Inspecting and Editing Knowledge Representations in Language Models
Evan Hernandez, Belinda Z. Li, Jacob Andreas
Neural language models (LMs) represent facts about the world described by text. Sometimes these facts derive from training data (in most LMs, a representation of the word "banana"…
LaMPP: Language Models as Probabilistic Priors for Perception and Action
Belinda Z. Li, William Chen, Pratyusha Sharma +1
Language models trained on large text corpora encode rich distributional information about real-world environments and action sequences. This information plays a crucial role in cu…
Language Modeling with Latent Situations
Belinda Z. Li, Maxwell Nye, Jacob Andreas
Language models (LMs) often generate incoherent outputs: they refer to events and entity states that are incompatible with the state of the world described in their inputs. We intr…
Quantifying Adaptability in Pre-trained Language Models with 500 Tasks
Belinda Z. Li, Jane Yu, Madian Khabsa +3
When a neural language model (LM) is adapted to perform a new task, what aspects of the task predict the eventual performance of the model? In NLP, systematic features of LM genera…
Implicit Representations of Meaning in Neural Language Models
Belinda Z. Li, Maxwell Nye, Jacob Andreas
Does the effectiveness of neural language models derive entirely from accurate modeling of surface word co-occurrence statistics, or do these models represent and reason about the…
On Unifying Misinformation Detection
Nayeon Lee, Belinda Z. Li, Sinong Wang +4
In this paper, we introduce UnifiedM2, a general-purpose misinformation model that jointly models multiple domains of misinformation with a single, unified setup. The model is trai…
On the Influence of Masking Policies in Intermediate Pre-training
Qinyuan Ye, Belinda Z. Li, Sinong Wang +5
Current NLP models are predominantly trained through a two-stage "pre-train then fine-tune" pipeline. Prior work has shown that inserting an intermediate pre-training stage, using…
Studying Strategically: Learning to Mask for Closed-book QA
Qinyuan Ye, Belinda Z. Li, Sinong Wang +5
Closed-book question-answering (QA) is a challenging task that requires a model to directly answer questions without access to external knowledge. It has been shown that directly f…
Efficient One-Pass End-to-End Entity Linking for Questions
Belinda Z. Li, Sewon Min, Srinivasan Iyer +2
We present ELQ, a fast end-to-end entity linking model for questions, which uses a biencoder to jointly perform mention detection and linking in one pass. Evaluated on WebQSP and G…
Language Models as Fact Checkers?
Nayeon Lee, Belinda Z. Li, Sinong Wang +3
Recent work has suggested that language models (LMs) store both common-sense and factual knowledge learned from pre-training data. In this paper, we leverage this implicit knowledg…