9 accepted papers
We investigate a family of simple long-context attacks on large language models: prompting with hundreds of demonstrations of undesirable behavior. This attack is newly feasible with the larger context windows recently deployed by language model providers like Google DeepMind, OpenAI and Anthropic.…
Prompting is the primary way to utilize the multitask capabilities of language models (LMs), but prompts occupy valuable space in the input context window, and repeatedly encoding the same prompt is computationally inefficient. Finetuning and distillation methods allow for specialization of LMs with…
Models can fail in unpredictable ways during deployment due to task ambiguity, when multiple behaviors are consistent with the provided training data. An example is an object classifier trained on red squares and blue circles: when encountering blue squares, the intended behavior is undefined. We in…
Reinforcement learning (RL) agents are particularly hard to train when rewards are sparse. One common solution is to use intrinsic rewards to encourage agents to explore their environment. However, recent intrinsic exploration methods often use state-based novelty measures which reward low-level exp…
Recent work has shown that augmenting environments with language descriptions improves policy learning. However, for environments with complex language abstractions, learning how to ground language to observations is difficult due to sparse, delayed rewards. We propose Language Dynamics Distillation…
Generating step-by-step "chain-of-thought" rationales improves language model performance on complex reasoning tasks like mathematics or commonsense question-answering. However, inducing language model rationale generation currently requires either constructing massive rationale datasets or sacrific…
To be good conversational partners, natural language processing (NLP) systems should be trained to produce contextually useful utterances. Prior work has investigated training NLP systems with communication-based objectives, where a neural listener stands in as a communication partner. However, thes…
To build agents that can collaborate effectively with others, recent research has trained artificial agents to communicate with each other in Lewis-style referential games. However, this often leads to successful but uninterpretable communication. We argue that this is due to the game objective: com…
We describe a procedure for explaining neurons in deep representations by identifying compositional logical concepts that closely approximate neuron behavior. Compared to prior work that uses atomic labels as explanations, analyzing neurons compositionally allows us to more precisely and expressivel…