Can foundation models actively gather information in interactive environments to test hypotheses?

• URL: http://arxiv.org/abs/2412.06438v2
• Date: Tue, 07 Oct 2025 11:28:32 GMT
• Title: Can foundation models actively gather information in interactive environments to test hypotheses?
• Authors: Danny P. Sawyer, Nan Rosemary Ke, Hubert Soyer, Martin Engelcke, David P Reichert, Drew A. Hudson, John Reid, Alexander Lerchner, Danilo Jimenez Rezende, Timothy P Lillicrap, Michael Mozer, Jane X Wang,
• Abstract summary: Foundation models excel at single-turn reasoning but struggle with multi-turn exploration in dynamic environments.<n>We evaluated these models on their ability to learn from experience, adapt, and gather information.
• Score: 43.42688356541211
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Foundation models excel at single-turn reasoning but struggle with multi-turn exploration in dynamic environments, a requirement for many real-world challenges. We evaluated these models on their ability to learn from experience, adapt, and gather information. First, in "Feature World," a simple setting for testing information gathering, models performed near-optimally. However, to test more complex, multi-trial learning, we implemented a text-based version of the "Alchemy" environment, a benchmark for meta-learning. Here, agents must deduce a latent causal structure by integrating information across many trials. In this setting, recent foundation models initially failed to improve their performance over time. Crucially, we found that prompting the models to summarize their observations at regular intervals enabled an emergent meta-learning process. This allowed them to improve across trials and even adaptively re-learn when the environment's rules changed unexpectedly. While most models handled the simple task, Alchemy revealed stark differences in robustness: Gemini 2.5 performed best, followed by Claude 3.7, while ChatGPT-4o and o4-mini struggled. This underscores Alchemy's value as a benchmark. Our findings demonstrate that the biggest challenge for foundation models is not selecting informative actions in the moment, but integrating knowledge through adaptive strategies over time. Encouragingly, there appears to be no intrinsic barrier to future models mastering these abilities.

Related papers

• Meta-Learning Transformers to Improve In-Context Generalization [8.694999451321571]
  In-context learning enables transformer models to generalize to new tasks based solely on input prompts.<n>Existing training paradigms typically rely on large, unstructured datasets that are costly to store.<n>We propose an alternative training strategy where we leverage a collection of multiple, small-scale, and domain-specific datasets.
  arXiv  Detail & Related papers  (2025-07-07T14:02:22Z)
• Do Retrieval-Augmented Language Models Adapt to Varying User Needs? [28.729041459278587]
  This paper introduces a novel evaluation framework that systematically assesses RALMs under three user need cases. By varying both user instructions and the nature of retrieved information, our approach captures the complexities of real-world applications. Our findings highlight the necessity of user-centric evaluations in the development of retrieval-augmented systems.
  arXiv  Detail & Related papers  (2025-02-27T05:39:38Z)
• Interactive Agents to Overcome Ambiguity in Software Engineering [61.40183840499932]
  AI agents are increasingly being deployed to automate tasks, often based on ambiguous and underspecified user instructions. Making unwarranted assumptions and failing to ask clarifying questions can lead to suboptimal outcomes. We study the ability of LLM agents to handle ambiguous instructions in interactive code generation settings by evaluating proprietary and open-weight models on their performance.
  arXiv  Detail & Related papers  (2025-02-18T17:12:26Z)
• SMamba: Sparse Mamba for Event-based Object Detection [17.141967728323714]
  Transformer-based methods have achieved remarkable performance in event-based object detection, owing to the global modeling ability. To mitigate cost, some researchers propose window attention based sparsification strategies to discard unimportant regions. We propose Sparse Mamba, which performs adaptive sparsification to reduce computational effort while maintaining global modeling ability.
  arXiv  Detail & Related papers  (2025-01-21T08:33:32Z)
• Dynamic Intelligence Assessment: Benchmarking LLMs on the Road to AGI with a Focus on Model Confidence [3.4049215220521933]
  We introduce Dynamic Intelligence Assessment (DIA), a novel methodology for testing AI models. The framework introduces four new metrics to assess a model's reliability and confidence across multiple attempts. The accompanying dataset, DIA-Bench, contains a diverse collection of challenge templates with mutable parameters presented in various formats.
  arXiv  Detail & Related papers  (2024-10-20T20:07:36Z)
• Adapting Vision-Language Models to Open Classes via Test-Time Prompt Tuning [50.26965628047682]
  Adapting pre-trained models to open classes is a challenging problem in machine learning. In this paper, we consider combining the advantages of both and come up with a test-time prompt tuning approach. Our proposed method outperforms all comparison methods on average considering both base and new classes.
  arXiv  Detail & Related papers  (2024-08-29T12:34:01Z)
• A Practitioner's Guide to Continual Multimodal Pretraining [83.63894495064855]
  Multimodal foundation models serve numerous applications at the intersection of vision and language.<n>To keep models updated, research into continual pretraining mainly explores scenarios with either infrequent, indiscriminate updates on large-scale new data, or frequent, sample-level updates.<n>We introduce FoMo-in-Flux, a continual multimodal pretraining benchmark with realistic compute constraints and practical deployment requirements.
  arXiv  Detail & Related papers  (2024-08-26T17:59:01Z)
• A Plug-and-Play Method for Rare Human-Object Interactions Detection by Bridging Domain Gap [50.079224604394]
  We present a novel model-agnostic framework called textbfContext-textbfEnhanced textbfFeature textbfAment (CEFA) CEFA consists of a feature alignment module and a context enhancement module. Our method can serve as a plug-and-play module to improve the detection performance of HOI models on rare categories.
  arXiv  Detail & Related papers  (2024-07-31T08:42:48Z)
• Test-Time Adaptation for Combating Missing Modalities in Egocentric Videos [92.38662956154256]
  Real-world applications often face challenges with incomplete modalities due to privacy concerns, efficiency needs, or hardware issues.<n>We propose a novel approach to address this issue at test time without requiring retraining.<n>MiDl represents the first self-supervised, online solution for handling missing modalities exclusively at test time.
  arXiv  Detail & Related papers  (2024-04-23T16:01:33Z)
• Relational Object-Centric Actor-Critic [44.99833362998488]
  Recent works highlight that disentangled object representations can aid policy learning in image-based, object-centric reinforcement learning tasks. This paper proposes a novel object-centric reinforcement learning algorithm that integrates actor-critic and model-based approaches. We evaluate our method in a simulated 3D robotic environment and a 2D environment with compositional structure.
  arXiv  Detail & Related papers  (2023-10-26T06:05:12Z)
• Active Sensing with Predictive Coding and Uncertainty Minimization [0.0]
  We present an end-to-end procedure for embodied exploration inspired by two biological computations. We first demonstrate our approach in a maze navigation task and show that it can discover the underlying transition distributions and spatial features of the environment. We show that our model builds unsupervised representations through exploration that allow it to efficiently categorize visual scenes.
  arXiv  Detail & Related papers  (2023-07-02T21:14:49Z)
• Discover, Explanation, Improvement: An Automatic Slice Detection Framework for Natural Language Processing [72.14557106085284]
  slice detection models (SDM) automatically identify underperforming groups of datapoints. This paper proposes a benchmark named "Discover, Explain, improve (DEIM)" for classification NLP tasks. Our evaluation shows that Edisa can accurately select error-prone datapoints with informative semantic features.
  arXiv  Detail & Related papers  (2022-11-08T19:00:00Z)
• Synthetic Model Combination: An Instance-wise Approach to Unsupervised Ensemble Learning [92.89846887298852]
  Consider making a prediction over new test data without any opportunity to learn from a training set of labelled data. Give access to a set of expert models and their predictions alongside some limited information about the dataset used to train them.
  arXiv  Detail & Related papers  (2022-10-11T10:20:31Z)
• Should Models Be Accurate? [14.044354912031864]
  We focus our investigations on Dyna-style planning in a prediction setting. We introduce a meta-learning algorithm for training models with a focus on their usefulness to the learner instead of their accuracy in modelling the environment. Our experiments show that our algorithm enables faster learning than even using an accurate model built with domain-specific knowledge of the non-stationarity.
  arXiv  Detail & Related papers  (2022-05-22T04:23:54Z)
• Exploring the Trade-off between Plausibility, Change Intensity and Adversarial Power in Counterfactual Explanations using Multi-objective Optimization [73.89239820192894]
  We argue that automated counterfactual generation should regard several aspects of the produced adversarial instances. We present a novel framework for the generation of counterfactual examples.
  arXiv  Detail & Related papers  (2022-05-20T15:02:53Z)
• Information-Theoretic Odometry Learning [83.36195426897768]
  We propose a unified information theoretic framework for learning-motivated methods aimed at odometry estimation. The proposed framework provides an elegant tool for performance evaluation and understanding in information-theoretic language.
  arXiv  Detail & Related papers  (2022-03-11T02:37:35Z)
• Sim2Real Object-Centric Keypoint Detection and Description [40.58367357980036]
  Keypoint detection and description play a central role in computer vision. We propose the object-centric formulation, which requires further identifying which object each interest point belongs to. We develop a sim2real contrastive learning mechanism that can generalize the model trained in simulation to real-world applications.
  arXiv  Detail & Related papers  (2022-02-01T15:00:20Z)
• Exploring Strategies for Generalizable Commonsense Reasoning with Pre-trained Models [62.28551903638434]
  We measure the impact of three different adaptation methods on the generalization and accuracy of models. Experiments with two models show that fine-tuning performs best, by learning both the content and the structure of the task, but suffers from overfitting and limited generalization to novel answers. We observe that alternative adaptation methods like prefix-tuning have comparable accuracy, but generalize better to unseen answers and are more robust to adversarial splits.
  arXiv  Detail & Related papers  (2021-09-07T03:13:06Z)
• Online reinforcement learning with sparse rewards through an active inference capsule [62.997667081978825]
  This paper introduces an active inference agent which minimizes the novel free energy of the expected future. Our model is capable of solving sparse-reward problems with a very high sample efficiency. We also introduce a novel method for approximating the prior model from the reward function, which simplifies the expression of complex objectives.
  arXiv  Detail & Related papers  (2021-06-04T10:03:36Z)
• Reinforcement Learning for Sparse-Reward Object-Interaction Tasks in a First-person Simulated 3D Environment [73.9469267445146]
  First-person object-interaction tasks in high-fidelity, 3D, simulated environments such as the AI2Thor pose significant sample-efficiency challenges for reinforcement learning agents. We show that one can learn object-interaction tasks from scratch without supervision by learning an attentive object-model as an auxiliary task.
  arXiv  Detail & Related papers  (2020-10-28T19:27:26Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.