Search Results for author:
Found 36 papers, 14 papers with code
Training language models to follow instructions with human feedback
In this paper, we show an avenue for aligning language models with user intent on a wide range of tasks by fine-tuning with human feedback.
GPT-4 Technical Report
We report the development of GPT-4, a large-scale, multimodal model which can accept image and text inputs and produce text outputs.
Ranked #1 on
Long-Context Understanding
on Ada-LEval (BestAnswer)
Evaluating Large Language Models Trained on Code
We introduce Codex, a GPT language model fine-tuned on publicly available code from GitHub, and study its Python code-writing capabilities.
Ranked #1 on
Multi-task Language Understanding
on BBH-alg
Let's Verify Step by Step
We conduct our own investigation, finding that process supervision significantly outperforms outcome supervision for training models to solve problems from the challenging MATH dataset.
Ranked #1 on
Math Word Problem Solving
on MATH minival
(using extra training data)
AI Safety Gridworlds
We present a suite of reinforcement learning environments illustrating various safety properties of intelligent agents.
Deep reinforcement learning from human preferences
For sophisticated reinforcement learning (RL) systems to interact usefully with real-world environments, we need to communicate complex goals to these systems.
Universal Reinforcement Learning Algorithms: Survey and Experiments
Many state-of-the-art reinforcement learning (RL) algorithms typically assume that the environment is an ergodic Markov Decision Process (MDP).
Generalised Discount Functions applied to a Monte-Carlo AImu Implementation
We have added to the GRL simulation platform AIXIjs the functionality to assign an agent arbitrary discount functions, and an environment which can be used to determine the effect of discounting on an agent's policy.
Reward learning from human preferences and demonstrations in Atari
To solve complex real-world problems with reinforcement learning, we cannot rely on manually specified reward functions.
Scalable agent alignment via reward modeling: a research direction
One obstacle to applying reinforcement learning algorithms to real-world problems is the lack of suitable reward functions.
Learning Human Objectives by Evaluating Hypothetical Behavior
To address this challenge, we propose an algorithm that safely and interactively learns a model of the user's reward function.
Quantifying Differences in Reward Functions
However, this method cannot distinguish between the learned reward function failing to reflect user preferences and the policy optimization process failing to optimize the learned reward.
Self-critiquing models for assisting human evaluators
On a topic-based summarization task, critiques written by our models help humans find flaws in summaries that they would have otherwise missed.
Learning to Understand Goal Specifications by Modelling Reward
Recent work has shown that deep reinforcement-learning agents can learn to follow language-like instructions from infrequent environment rewards.
Nonparametric General Reinforcement Learning
However, there are Bayesian approaches to general RL that satisfy objective optimality guarantees: We prove that Thompson sampling is asymptotically optimal in stochastic environments in the sense that its value converges to the value of the optimal policy.
Exploration Potential
We introduce exploration potential, a quantity that measures how much a reinforcement learning agent has explored its environment class.
A Formal Solution to the Grain of Truth Problem
In this paper we present a formal and general solution to the full grain of truth problem: we construct a class of policies that contains all computable policies as well as Bayes-optimal policies for every lower semicomputable prior over the class.
Thompson Sampling is Asymptotically Optimal in General Environments
We discuss a variant of Thompson sampling for nonparametric reinforcement learning in a countable classes of general stochastic environments.
Loss Bounds and Time Complexity for Speed Priors
On a polynomial time computable sequence our speed prior is computable in exponential time.
On the Computability of AIXI
Solomonoff induction and the reinforcement learning agent AIXI are proposed answers to this question.
Bad Universal Priors and Notions of Optimality
A big open question of algorithmic information theory is the choice of the universal Turing machine (UTM).
On the Computability of Solomonoff Induction and Knowledge-Seeking
Solomonoff induction is held as a gold standard for learning, but it is known to be incomputable.
Solomonoff Induction Violates Nicod's Criterion
Nicod's criterion states that observing a black raven is evidence for the hypothesis H that all ravens are black.
Sequential Extensions of Causal and Evidential Decision Theory
Moving beyond the dualistic view in AI where agent and environment are separated incurs new challenges for decision making, as calculation of expected utility is no longer straightforward.
A Definition of Happiness for Reinforcement Learning Agents
What is happiness for reinforcement learning agents?
Indefinitely Oscillating Martingales
We construct a class of nonnegative martingale processes that oscillate indefinitely with high probability.
Scaling shared model governance via model splitting
Currently the only techniques for sharing governance of a deep learning model are homomorphic encryption and secure multiparty computation.
Pitfalls of learning a reward function online
We formally introduce two desirable properties: the first is `unriggability', which prevents the agent from steering the learning process in the direction of a reward function that is easier to optimise.
Hidden Incentives for Auto-Induced Distributional Shift
We introduce the term auto-induced distributional shift (ADS) to describe the phenomenon of an algorithm causing a change in the distribution of its own inputs.
Active Reinforcement Learning: Observing Rewards at a Cost
Active reinforcement learning (ARL) is a variant on reinforcement learning where the agent does not observe the reward unless it chooses to pay a query cost c > 0.
Institutionalising Ethics in AI through Broader Impact Requirements
In 2020, the Conference on Neural Information Processing Systems (NeurIPS) introduced a requirement for submitting authors to include a statement on the broader societal impacts of their research.
Recursively Summarizing Books with Human Feedback
Our human labelers are able to supervise and evaluate the models quickly, despite not having read the entire books themselves.
Revealing the Incentive to Cause Distributional Shift
We use these unit tests to demonstrate that changes to the learning algorithm (e. g. introducing meta-learning) can cause previously hidden incentives to be revealed, resulting in qualitatively different behaviour despite no change in performance metric.
Hidden incentives for self-induced distributional shift
Decisions made by machine learning systems have increasing influence on the world.
Safe Deep RL in 3D Environments using Human Feedback
In this paper we answer this question in the affirmative, using ReQueST to train an agent to perform a 3D first-person object collection task using data entirely from human contractors.
Weak-to-Strong Generalization: Eliciting Strong Capabilities With Weak Supervision
Widely used alignment techniques, such as reinforcement learning from human feedback (RLHF), rely on the ability of humans to supervise model behavior - for example, to evaluate whether a model faithfully followed instructions or generated safe outputs.
