No file available
To read the file of this research,
you can request a copy directly from the authors.
Passive learning of active causal strategies in agents and language models
Preprints and early-stage research may not have been peer reviewed yet.
Abstract
What can be learned about causality and experimentation from passive data? This question is salient given recent successes of passively-trained language models in interactive domains such as tool use. Passive learning is inherently limited. However, we show that purely passive learning can in fact allow an agent to learn generalizable strategies for determining and using causal structures, as long as the agent can intervene at test time. We formally illustrate that learning a strategy of first experimenting, then seeking goals, can allow generalization from passive learning in principle. We then show empirically that agents trained via imitation on expert data can indeed generalize at test time to infer and use causal links which are never present in the training data; these agents can also generalize experimentation strategies to novel variable sets never observed in training. We then show that strategies for causal intervention and exploitation can be generalized from passive data even in a more complex environment with high-dimensional observations, with the support of natural language explanations. Explanations can even allow passive learners to generalize out-of-distribution from perfectly-confounded training data. Finally, we show that language models, trained only on passive next-word prediction, can generalize causal intervention strategies from a few-shot prompt containing examples of experimentation, together with explanations and reasoning. These results highlight the surprising power of passive learning of active causal strategies, and may help to understand the behaviors and capabilities of language models.
ResearchGate has not been able to resolve any citations for this publication.
Importance:
A previous meta-analysis of the association between alcohol use and all-cause mortality found no statistically significant reductions in mortality risk at low levels of consumption compared with lifetime nondrinkers. However, the risk estimates may have been affected by the number and quality of studies then available, especially those for women and younger cohorts.
Objective:
To investigate the association between alcohol use and all-cause mortality, and how sources of bias may change results.
Data sources:
A systematic search of PubMed and Web of Science was performed to identify studies published between January 1980 and July 2021.
Study selection:
Cohort studies were identified by systematic review to facilitate comparisons of studies with and without some degree of controls for biases affecting distinctions between abstainers and drinkers. The review identified 107 studies of alcohol use and all-cause mortality published from 1980 to July 2021.
Data extraction and synthesis:
Mixed linear regression models were used to model relative risks, first pooled for all studies and then stratified by cohort median age (<56 vs ≥56 years) and sex (male vs female). Data were analyzed from September 2021 to August 2022.
Main outcomes and measures:
Relative risk estimates for the association between mean daily alcohol intake and all-cause mortality.
Results:
There were 724 risk estimates of all-cause mortality due to alcohol intake from the 107 cohort studies (4 838 825 participants and 425 564 deaths available) for the analysis. In models adjusting for potential confounding effects of sampling variation, former drinker bias, and other prespecified study-level quality criteria, the meta-analysis of all 107 included studies found no significantly reduced risk of all-cause mortality among occasional (>0 to <1.3 g of ethanol per day; relative risk [RR], 0.96; 95% CI, 0.86-1.06; P = .41) or low-volume drinkers (1.3-24.0 g per day; RR, 0.93; P = .07) compared with lifetime nondrinkers. In the fully adjusted model, there was a nonsignificantly increased risk of all-cause mortality among drinkers who drank 25 to 44 g per day (RR, 1.05; P = .28) and significantly increased risk for drinkers who drank 45 to 64 and 65 or more grams per day (RR, 1.19 and 1.35; P < .001). There were significantly larger risks of mortality among female drinkers compared with female lifetime nondrinkers (RR, 1.22; P = .03).
Conclusions and relevance:
In this updated systematic review and meta-analysis, daily low or moderate alcohol intake was not significantly associated with all-cause mortality risk, while increased risk was evident at higher consumption levels, starting at lower levels for women than men.
What do we communicate with causal explanations? Upon being told, "E because C", a person might learn that C and E both occurred, and perhaps that there is a causal relationship between C and E. In fact, causal explanations systematically disclose much more than this basic information. Here, we offer a communication-theoretic account of explanation that makes specific predictions about the kinds of inferences people draw from others' explanations. We test these predictions in a case study involving the role of norms and causal structure. In Experiment 1, we demonstrate that people infer the normality of a cause from an explanation when they know the underlying causal structure. In Experiment 2, we show that people infer the causal structure from an explanation if they know the normality of the cited cause. We find these patterns both for scenarios that manipulate the statistical and prescriptive normality of events. Finally, we consider how the communicative function of explanations, as highlighted in this series of experiments, may help to elucidate the distinctive roles that normality and causal structure play in causal judgment, paving the way toward a more comprehensive account of causal explanation. (PsycInfo Database Record (c) 2021 APA, all rights reserved).
A fundamental task in various disciplines of science, including biology, is to find underlying causal relations and make use of them. Causal relations can be seen if interventions are properly applied; however, in many cases they are difficult or even impossible to conduct. It is then necessary to discover causal relations by analyzing statistical properties of purely observational data, which is known as causal discovery or causal structure search. This paper aims to give a introduction to and a brief review of the computational methods for causal discovery that were developed in the past three decades, including constraint-based and score-based methods and those based on functional causal models, supplemented by some illustrations and applications.
Recent approaches in robot perception follow the insight that perception is facilitated by interaction with the environment. These approaches are subsumed under the term Interactive Perception (IP). This view of perception provides the following benefits. First, interaction with the environment creates a rich sensory signal that would otherwise not be present. Second, knowledge of the regularity in the combined space of sensory data and action parameters facilitates the prediction and interpretation of the sensory signal. In this survey, we postulate this as a principle for robot perception and collect evidence in its support by analyzing and categorizing existing work in this area. We also provide an overview of the most important applications of IP. We close this survey by discussing remaining open questions. With this survey, we hope to help define the field of Interactive Perception and to provide a valuable resource for future research.
Recent progress in artificial intelligence (AI) has renewed interest in building systems that learn and think like people. Many advances have come from using deep neural networks trained end-to-end in tasks such as object recognition, video games, and board games, achieving performance that equals or even beats humans in some respects. Despite their biological inspiration and performance achievements, these systems differ from human intelligence in crucial ways. We review progress in cognitive science suggesting that truly human-like learning and thinking machines will have to reach beyond current engineering trends in both what they learn, and how they learn it. Specifically, we argue that these machines should (a) build causal models of the world that support explanation and understanding, rather than merely solving pattern recognition problems; (b) ground learning in intuitive theories of physics and psychology, to support and enrich the knowledge that is learned; and (c) harness compositionality and learning-to-learn to rapidly acquire and generalize knowledge to new tasks and situations. We suggest concrete challenges and promising routes towards these goals that can combine the strengths of recent neural network advances with more structured cognitive models.
People can test hypotheses through either selection or reception. In a selection task, the learner actively chooses observations to test his or her beliefs, whereas in reception tasks data are passively encountered. People routinely use both forms of testing in everyday life, but the critical psychological differences between selection and reception learning remain poorly understood. One hypothesis is that selection learning improves learning performance by enhancing generic cognitive processes related to motivation, attention, and engagement. Alternatively, we suggest that differences between these 2 learning modes derives from a hypothesis-dependent sampling bias that is introduced when a person collects data to test his or her own individual hypothesis. Drawing on influential models of sequential hypothesis-testing behavior, we show that such a bias (a) can lead to the collection of data that facilitates learning compared with reception learning and (b) can be more effective than observing the selections of another person. We then report a novel experiment based on a popular category learning paradigm that compares reception and selection learning. We additionally compare selection learners to a set of "yoked" participants who viewed the exact same sequence of observations under reception conditions. The results revealed systematic differences in performance that depended on the learner's role in collecting information and the abstract structure of the problem. (PsycINFO Database Record (c) 2013 APA, all rights reserved).
We consider structural equation models in which variables can be written as a function of their parents and noise terms, which
are assumed to be jointly independent. Corresponding to each structural equation model is a directed acyclic graph describing
the relationships between the variables. In Gaussian structural equation models with linear functions, the graph can be identified
from the joint distribution only up to Markov equivalence classes, assuming faithfulness. In this work, we prove full identifiability
in the case where all noise variables have the same variance: the directed acyclic graph can be recovered from the joint Gaussian
distribution. Our result has direct implications for causal inference: if the data follow a Gaussian structural equation model
with equal error variances, then, assuming that all variables are observed, the causal structure can be inferred from observational
data only. We propose a statistical method and an algorithm based on our theoretical findings.
Full and exact adaptation to sensory rearrangement in adult human Ss requires movement-produced sensory feedback. Riesen's work suggested that this factor also operates in the development of higher mammals but he proposed that sensory-sensory associations are the proposed that sensory-sensory associations are the prerequisite. To test these alternatives, visual stimulation of the active member (A) of each of 10 pairs of neonatal kittens was allowed to vary with its locomotor movements while equivalent stimulation of the second member (P) resulted from passive motion. Subsequent tests of visually guided paw placement, discrimination on a visual cliff, and the blink response were normal for A but failing in P. When other alternative explanations are excluded, this result extends the conclusions of studies of adult rearrangement to neonatal development. (18 ref.)
Exploratory activities seem to be intrinsically rewarding for children and crucial for their cognitive development. Can a machine be endowed with such an intrinsic motivation system? This is the question we study in this paper, presenting a number of computational systems that try to capture this drive towards novel or curious situations. After discussing related research coming from developmental psychology, neuroscience, developmental robotics, and active learning, this paper presents the mechanism of Intelligent Adaptive Curiosity, an intrinsic motivation system which pushes a robot towards situations in which it maximizes its learning progress. This drive makes the robot focus on situations which are neither too predictable nor too unpredictable, thus permitting autonomous mental development. The complexity of the robot's activities autonomously increases and complex developmental sequences self-organize without being constructed in a supervised manner. Two experiments are presented illustrating the stage-like organization emerging with this mechanism. In one of them, a physical robot is placed on a baby play mat with objects that it can learn to manipulate. Experimental results show that the robot first spends time in situations which are easy to learn, then shifts its attention progressively to situations of increasing difficulty, avoiding situations in which nothing can be learned. Finally, these various results are discussed in relation to more complex forms of behavioral organization and data coming from developmental psychology
This paper surveys and organizes research works in a new paradigm in natural language processing, which we dub “prompt-based learning”. Unlike traditional supervised learning, which trains a model to take in an input x and predict an output y as P (y|x), prompt-based learning is based on language models that model the probability of text directly. To use these models to perform prediction tasks, the original input x is modified using a template into a textual string prompt x′ that has some unfilled slots, and then the language model is used to probabilistically fill the unfilled information to obtain a final string , from which the final output y can be derived. This framework is powerful and attractive for a number of reasons: it allows the language model to be pre-trained on massive amounts of raw text, and by defining a new prompting function the model is able to perform few-shot or even zero-shot learning, adapting to new scenarios with few or no labeled data. In this paper we introduce the basics of this promising paradigm, describe a unified set of mathematical notations that can cover a wide variety of existing work, and organize existing work along several dimensions, e.g. the choice of pre-trained language models, prompts, and tuning strategies. To make the field more accessible to interested beginners, we not only make a systematic review of existing works and a highly structured typology of prompt-based concepts, but also release other resources, e.g., a website NLPedia–Pretrain including constantly-updated survey, and paperlist.
Introduction
Studies evaluating alcohol consumption and cardiovascular diseases have shown inconsistent results.
Methods
We performed a systematic review of peer-reviewed publications from an extensive query of Ovid MEDLINE, Ovid Embase, Ovid Cochrane Database of Systematic Reviews, Scopus, and Web of Science from database inception to March 2022 for all studies that reported the association between alcohol consumption in terms of quantity (daily or weekly amounts) and type of beverage (wine, beer or spirit) and cardiovascular disease events.
Results
The study population included a total of 1,579,435 individuals based on 56 cohorts from several countries. We found that moderate wine consumption defined as 1-4 drinks per week was associated with a reduction in risk for cardiovascular mortality when compared to beer or spirits. However, higher risk for cardiovascular disease mortality was typically seen with heavier daily or weekly alcohol consumption across all types of beverages.
Conclusions
It is possible that the observational studies may overestimate the benefits of alcohol for cardiovascular disease outcomes. Although moderate wine consumption is probably associated with low cardiovascular disease events, there are many confounding factors, in particular lifestyle, genetic and socioeconomic associations with wine drinking which likely explain much of the association with wine and reduced cardiovascular disease events. Further prospective study of alcohol and all-cause mortality, including cancer, is needed.
We show that if any number of variables are allowed to be simultaneously and independently randomized in any one experiment, log_2(N) + 1 experiments are sufficient and in the worst case necessary to determine the causal relations among N ≥ 2 variables when no latent variables, no sample selection bias and no feedback cycles are present. For all K, 0 < K < 1/2 N we provide an upper bound on the number experiments required to determine causal structure when each experiment simultaneously randomizes K variables. For large N, these bounds are significantly lower than the N - 1 bound required when each experiment randomizes at most one variable. For k_(max) < N/2, we show that (N/k_(max) -1) + N/2k_(max) log_2(k_(max)) experiments are sufficient and in the worst case necessary. We offer a conjecture as to the minimal number of experiments that are in the worst case sufficient to identify all causal relations among N observed variables that are a subset of the vertices of a DAG.
THE DRAMATIC SUCCESS In machine learning has led to an explosion of artificial intelligence (AI) applications and increasing expectations for autonomous systems that exhibit human-level intelligence. These expectations have, however, met with fundamental obstacles that cut across many application areas. One such obstacle is adaptability, or robustness. Machine learning researchers have noted current systems lack the ability to recognize or react to new circumstances they have not been specifically programmed or trained for.
Current machine learning systems operate, almost exclusively, in a statistical, or model-blind mode, which entails severe theoretical limits on their power and performance. Such systems cannot reason about interventions and retrospection and, therefore, cannot serve as the basis for strong AI. To achieve human level intelligence, learning machines need the guidance of a model of reality, similar to the ones used in causal inference. To demonstrate the essential role of such models, I will present a summary of seven tasks which are beyond reach of current machine learning systems and which have been accomplished using the tools of causal inference.
This review presents empirical researchers with recent advances in causal inference, and stresses the paradigmatic shifts that must be un-dertaken in moving from traditional statistical analysis to causal analysis of multivariate data. Special emphasis is placed on the assumptions that un-derly all causal inferences, the languages used in formulating those assump-tions, the conditional nature of all causal and counterfactual claims, and the methods that have been developed for the assessment of such claims. These advances are illustrated using a general theory of causation based on the Structural Causal Model (SCM) described in Pearl (2000a), which subsumes and unifies other approaches to causation, and provides a coher-ent mathematical foundation for the analysis of causes and counterfactuals. In particular, the paper surveys the development of mathematical tools for inferring (from a combination of data and assumptions) answers to three types of causal queries: (1) queries about the effects of potential interven-tions, (also called "causal effects" or "policy evaluation") (2) queries about probabilities of counterfactuals, (including assessment of "regret," "attri-bution" or "causes of effects") and (3) queries about direct and indirect effects (also known as "mediation"). Finally, the paper defines the formal and conceptual relationships between the structural and potential-outcome frameworks and presents tools for a symbiotic analysis that uses the strong features of both.
Teleological explanations (TEs) account for the existence or properties of an entity in terms of a function: we have hearts because they pump blood, and telephones for communication. While many teleological explanations seem appropriate, others are clearly not warranted--for example, that rain exists for plants to grow. Five experiments explore the theoretical commitments that underlie teleological explanations. With the analysis of [Wright, L. (1976). Teleological Explanations. Berkeley, CA: University of California Press] from philosophy as a point of departure, we examine in Experiment 1 whether teleological explanations are interpreted causally, and confirm that TEs are only accepted when the function invoked in the explanation played a causal role in bringing about what is being explained. However, we also find that playing a causal role is not sufficient for all participants to accept TEs. Experiment 2 shows that this is not because participants fail to appreciate the causal structure of the scenarios used as stimuli. In Experiments 3-5 we show that the additional requirement for TE acceptance is that the process by which the function played a causal role must be general in the sense of conforming to a predictable pattern. These findings motivate a proposal, Explanation for Export, which suggests that a psychological function of explanation is to highlight information likely to subserve future prediction and intervention. We relate our proposal to normative accounts of explanation from philosophy of science, as well as to claims from psychology and artificial intelligence.
Generating and evaluating explanations is spontaneous, ubiquitous and fundamental to our sense of understanding. Recent evidence suggests that in the course of an individual's reasoning, engaging in explanation can have profound effects on the probability assigned to causal claims, on how properties are generalized and on learning. These effects follow from two properties of the structure of explanations: explanations accommodate novel information in the context of prior beliefs, and do so in a way that fosters generalization. The study of explanation thus promises to shed light on core cognitive issues, such as learning, induction and conceptual representation. Moreover, the influence of explanation on learning and inference presents a challenge to theories that neglect the roles of prior knowledge and explanation-based reasoning.
An optimistic perspective on offline reinforcement learning
- Rishabh Agarwal
- Dale Schuurmans
- Mohammad Norouzi
Rishabh Agarwal, Dale Schuurmans, and Mohammad Norouzi. An optimistic perspective
on offline reinforcement learning. In International Conference on Machine Learning, pages
104-114. PMLR, 2020.
All you need is supervised learning: From imitation learning to meta-rl with upside down rl
- Kai Arulkumaran
- R Dylan
- Jürgen Ashley
- Rupesh K Schmidhuber
- Srivastava
Kai Arulkumaran, Dylan R Ashley, Jürgen Schmidhuber, and Rupesh K Srivastava. All you
need is supervised learning: From imitation learning to meta-rl with upside down rl. arXiv
preprint arXiv:2202.11960, 2022.
JAX: composable transformations of Python+NumPy programs
- James Bradbury
- Roy Frostig
- Peter Hawkins
- Matthew James Johnson
- Chris Leary
- Dougal Maclaurin
- George Necula
- Adam Paszke
- Jake Vanderplas
- Skye Wanderman-Milne
- Qiao Zhang
James Bradbury, Roy Frostig, Peter Hawkins, Matthew James Johnson, Chris Leary, Dougal
Maclaurin, George Necula, Adam Paszke, Jake VanderPlas, Skye Wanderman-Milne, and
Qiao Zhang. JAX: composable transformations of Python+NumPy programs, 2018. URL
http://github.com/google/jax.
Grounding large language models in interactive environments with online reinforcement learning
- Thomas Carta
- Clément Romac
- Thomas Wolf
- Sylvain Lamprier
- Olivier Sigaud
- Pierre-Yves Oudeyer
Thomas Carta, Clément Romac, Thomas Wolf, Sylvain Lamprier, Olivier Sigaud, and Pierre-Yves Oudeyer. Grounding large language models in interactive environments with online
reinforcement learning. arXiv preprint arXiv:2302.02662, 2023.
Transformers generalize differently from information stored in context vs in weights
- C Y Stephanie
- Ishita Chan
- Junkyung Dasgupta
- Dharshan Kim
- Kumaran
- Felix Andrew K Lampinen
- Hill
Stephanie CY Chan, Ishita Dasgupta, Junkyung Kim, Dharshan Kumaran, Andrew K Lampinen,
and Felix Hill. Transformers generalize differently from information stored in context vs in
weights. MemARI workshop, NeurIPS, 2022.
Data distributional properties drive emergent few-shot learning in transformers
- C Y Stephanie
- Adam Chan
- Santoro
- Jane X Andrew K Lampinen
- Aaditya Wang
- Singh
- H Pierre
- Jay Richemond
- Felix Mcclelland
- Hill
Stephanie CY Chan, Adam Santoro, Andrew K Lampinen, Jane X Wang, Aaditya Singh,
Pierre H Richemond, Jay McClelland, and Felix Hill. Data distributional properties drive
emergent few-shot learning in transformers. Advances in Neural Information Processing
Systems, 2022.
Aravind Srinivas, and Igor Mordatch. Decision transformer: Reinforcement learning via sequence modeling
- Lili Chen
- Kevin Lu
- Aravind Rajeswaran
- Kimin Lee
- Aditya Grover
- Misha Laskin
- Pieter Abbeel
Lili Chen, Kevin Lu, Aravind Rajeswaran, Kimin Lee, Aditya Grover, Misha Laskin, Pieter
Abbeel, Aravind Srinivas, and Igor Mordatch. Decision transformer: Reinforcement learning
via sequence modeling. Advances in neural information processing systems, 34:15084-15097,
2021.
Deep reinforcement learning from human preferences. Advances in neural information processing systems
- Jan Paul F Christiano
- Tom Leike
- Miljan Brown
- Shane Martic
- Dario Legg
- Amodei
Paul F Christiano, Jan Leike, Tom Brown, Miljan Martic, Shane Legg, and Dario Amodei. Deep
reinforcement learning from human preferences. Advances in neural information processing
systems, 30, 2017.
- Ishita Dasgupta
- Jane Wang
- Silvia Chiappa
- Jovana Mitrovic
- Pedro Ortega
- David Raposo
- Edward Hughes
- Peter Battaglia
- Matthew Botvinick
- Zeb Kurth-Nelson
Ishita Dasgupta, Jane Wang, Silvia Chiappa, Jovana Mitrovic, Pedro Ortega, David Raposo,
Edward Hughes, Peter Battaglia, Matthew Botvinick, and Zeb Kurth-Nelson. Causal reasoning
from meta-reinforcement learning. arXiv preprint arXiv:1901.08162, 2019.
Causal confusion in imitation learning
- Dinesh Pim De Haan
- Sergey Jayaraman
- Levine
Pim De Haan, Dinesh Jayaraman, and Sergey Levine. Causal confusion in imitation learning.
Advances in Neural Information Processing Systems, 32, 2019.
Generalizing goal-conditioned reinforcement learning with variational causal reasoning
- Wenhao Ding
- Haohong Lin
- Bo Li
- Ding Zhao
Wenhao Ding, Haohong Lin, Bo Li, and Ding Zhao. Generalizing goal-conditioned reinforcement learning with variational causal reasoning. arXiv preprint arXiv:2207.09081, 2022.
Off-policy deep reinforcement learning without exploration
- Scott Fujimoto
- David Meger
- Doina Precup
Scott Fujimoto, David Meger, and Doina Precup. Off-policy deep reinforcement learning
without exploration. In Kamalika Chaudhuri and Ruslan Salakhutdinov, editors, Proceedings of
the 36th International Conference on Machine Learning, volume 97 of Proceedings of Machine
Learning Research, pages 2052-2062. PMLR, 09-15 Jun 2019. URL https://proceedings.
mlr.press/v97/fujimoto19a.html.
Self-directed learning: A cognitive and computational perspective
- M Todd
- Douglas B Gureckis
- Markant
Todd M Gureckis and Douglas B Markant. Self-directed learning: A cognitive and computational perspective. Perspectives on Psychological Science, 7(5):464-481, 2012.
Haiku: Sonnet for JAX
- Tom Hennigan
- Trevor Cai
- Tamara Norman
- Igor Babuschkin
Tom Hennigan, Trevor Cai, Tamara Norman, and Igor Babuschkin. Haiku: Sonnet for JAX,
2020. URL http://github.com/deepmind/dm-haiku.
Training compute-optimal large language models
- Jordan Hoffmann
- Sebastian Borgeaud
- Arthur Mensch
- Elena Buchatskaya
- Trevor Cai
- Eliza Rutherford
- Diego De Las
- Lisa Anne Casas
- Johannes Hendricks
- Aidan Welbl
- Clark
Jordan Hoffmann, Sebastian Borgeaud, Arthur Mensch, Elena Buchatskaya, Trevor Cai, Eliza
Rutherford, Diego de Las Casas, Lisa Anne Hendricks, Johannes Welbl, Aidan Clark, et al.
Training compute-optimal large language models. arXiv preprint arXiv:2203.15556, 2022.
The curious case of neural text degeneration
- Ari Holtzman
- Jan Buys
- Li Du
- Maxwell Forbes
- Yejin Choi
Ari Holtzman, Jan Buys, Li Du, Maxwell Forbes, and Yejin Choi. The curious case of neural
text degeneration. In International Conference on Learning Representations, 2020.
Learning how to infer partial mdps for in-context adaptation and exploration
- Chentian Jiang
- Nan Rosemary Ke
- Hado Van Hasselt
Chentian Jiang, Nan Rosemary Ke, and Hado van Hasselt. Learning how to infer partial mdps
for in-context adaptation and exploration. arXiv preprint arXiv:2302.04250, 2023.
- Nan Rosemary Ke
- Silvia Chiappa
- Jane Wang
- Jorg Bornschein
- Theophane Weber
- Anirudh Goyal
- Matthew Botvinic
- Michael Mozer
- Danilo Jimenez Rezende
Nan Rosemary Ke, Silvia Chiappa, Jane Wang, Jorg Bornschein, Theophane Weber, Anirudh
Goyal, Matthew Botvinic, Michael Mozer, and Danilo Jimenez Rezende. Learning to induce
causal structure. arXiv preprint arXiv:2204.04875, 2022.
- Emre Kıcıman
- Robert Ness
- Amit Sharma
- Chenhao Tan
Emre Kıcıman, Robert Ness, Amit Sharma, and Chenhao Tan. Causal reasoning and large
language models: Opening a new frontier for causality. arXiv preprint arXiv:2305.00050, 2023.
Adam: A method for stochastic optimization
- P Diederik
- Jimmy Kingma
- Ba
Diederik P Kingma and Jimmy Ba. Adam: A method for stochastic optimization. arXiv preprint
arXiv:1412.6980, 2014.
Large language models are zero-shot reasoners
- Takeshi Kojima
- Shane Shixiang
- Machel Gu
- Yutaka Reid
- Yusuke Matsuo
- Iwasawa
Takeshi Kojima, Shixiang Shane Gu, Machel Reid, Yutaka Matsuo, and Yusuke Iwasawa. Large
language models are zero-shot reasoners. arXiv preprint arXiv:2205.11916, 2022.
Towards understanding how machines can learn causal overhypotheses
- Eliza Kosoy
- M David
- Adrian Chan
- Jasmine Liu
- Bryanna Collins
- Sandy Han Kaufmann
- Jessica B Huang
- John Hamrick
- Canny
Eliza Kosoy, David M Chan, Adrian Liu, Jasmine Collins, Bryanna Kaufmann, Sandy Han
Huang, Jessica B Hamrick, John Canny, Nan Rosemary Ke, and Alison Gopnik. Towards
understanding how machines can learn causal overhypotheses. arXiv preprint arXiv:2206.08353,
2022.
Stabilizing offpolicy q-learning via bootstrapping error reduction
- Aviral Kumar
- Justin Fu
- Matthew Soh
- George Tucker
- Sergey Levine
Aviral Kumar, Justin Fu, Matthew Soh, George Tucker, and Sergey Levine. Stabilizing offpolicy q-learning via bootstrapping error reduction. Advances in Neural Information Processing
Systems, 32, 2019.
Can language models learn from explanations in context? Findings of EMNLP
- Ishita Andrew K Lampinen
- Dasgupta
- C Y Stephanie
- Kory Chan
- Michael Henry Matthewson
- Antonia Tessler
- Creswell
- L James
- Jane X Mcclelland
- Felix Wang
- Hill
Andrew K Lampinen, Ishita Dasgupta, Stephanie CY Chan, Kory Matthewson, Michael Henry
Tessler, Antonia Creswell, James L McClelland, Jane X Wang, and Felix Hill. Can language
models learn from explanations in context? Findings of EMNLP, 2022.
Tell me why! explanations support learning relational and causal structure
- Nicholas Andrew K Lampinen
- Ishita Roy
- Dasgupta
- C Y Stephanie
- Allison Chan
- James Tam
- Chen Mcclelland
- Adam Yan
- Santoro
- C Neil
- Jane Rabinowitz
- Wang
Andrew K Lampinen, Nicholas Roy, Ishita Dasgupta, Stephanie CY Chan, Allison Tam, James
Mcclelland, Chen Yan, Adam Santoro, Neil C Rabinowitz, Jane Wang, et al. Tell me why!
explanations support learning relational and causal structure. In International Conference on
Machine Learning, pages 11868-11890. PMLR, 2022.
- Michael Laskin
- Luyu Wang
- Junhyuk Oh
- Emilio Parisotto
- Stephen Spencer
- Richie Steigerwald
- Strouse
Michael Laskin, Luyu Wang, Junhyuk Oh, Emilio Parisotto, Stephen Spencer, Richie Steigerwald, DJ Strouse, Steven Hansen, Angelos Filos, Ethan Brooks, et al. In-context reinforcement
learning with algorithm distillation. arXiv preprint arXiv:2210.14215, 2022.
Rectifier nonlinearities improve neural network acoustic models
- L Andrew
- Maas
- Y Awni
- Andrew Y Hannun
- Ng
Andrew L Maas, Awni Y Hannun, Andrew Y Ng, et al. Rectifier nonlinearities improve neural
network acoustic models. In Proc. icml, volume 30, page 3. Atlanta, Georgia, USA, 2013.
Causal induction from visual observations for goal directed tasks
- Suraj Nair
- Yuke Zhu
- Silvio Savarese
- Li Fei-Fei
Suraj Nair, Yuke Zhu, Silvio Savarese, and Li Fei-Fei. Causal induction from visual observations
for goal directed tasks. arXiv preprint arXiv:1910.01751, 2019.
Show and tell: Learning causal structures from observations and explanations
- Andrew Nam
- Christopher Hughes
- Thomas Icard
- Tobias Gerstenberg
Andrew Nam, Christopher Hughes, Thomas Icard, and Tobias Gerstenberg. Show and tell:
Learning causal structures from observations and explanations.