AGI Seminar Series

Awesome X

Interactive Embodied Agents


2405 - Vision-based Manipulation from Single Human Video with Open-World Object Graphs
2405 - From LLMs to Actions: Latent Codes as Bridges in Hierarchical Robot Control
2307 - Breadcrumbs to the Goal: Goal-Conditioned Exploration from Human-in-the-Loop Feedback
2210 - Language-Table: Interactive Language - Talking to Robots in Real Time


2403 - MineDreamer: Learning to Follow Instructions via Chain-of-Imagination for Simulated-World Control
2312 - MP5: A Multi-modal Open-ended Embodied System in Minecraft via Active Perception


Position Papers

2402 - Compositional Generative Modeling - A Single Model is Not All You Need

Composable World Models

2404 - Compete and Compose: Learning Independent Mechanisms for Modular World Models
20xx - Compositional Visual Generation with Energy Based Models [Igor Mordarch]

Backbone Architectures (Compositional Transformer)

2405 - Transformers Can Do Arithmetic with the Right Embeddings
2405 - Grokked Transformers are Implicit Reasoners: A Mechanistic Journey to the Edge of Generalization
2405 - Improving Transformers with Dynamically Composable Multi-Head Attention
2405 - Sparo: Selective Attention for Robust and Compositional Transformer Encodings for Vision
2404 - When Can Transformers Reason with Abstract Symbols?
2311 - Compositional Capabilities of Autoregressive Transformers: A Study on Synthetic, Interpretable Tasks
2311 - What Algorithms can Transformers Learn? A Study in Length Generalization
22xx - Making Transformers Solve Compositional Tasks
2110 - Compositional Attention: Disentangling Search and Retrieval


2405 - Neural Language of Thought Models
2312 - Reusable Slotwise Mechanisms
2310 - Object-Centric Architectures Enable Efficient Causal Representation Learning
2205 - HOWM - Toward Compositional Generalization in Object-Oriented World Modeling (ICML22)

In Human

2304 - Constructing Future Behaviour in The Hippocampal Formation Through Composition and Replay
2305 - Neural Knowledge Assembly in Humans and Neural Networks [Christopher Summerfield]
2304 - Hippocampal Spatio-Predictive Cognitive Maps Adaptively Guide Reward Generalization
2010 - Meta-Learning of Compositional Task Distributions in Humans and Machines [Thomas L. Griffiths]

Causal Representations

Position & Survey Papers

2403 - Large Language Models and Causal Inference in Collaboration: A Comprehensive Survey
2402 - Essential Role of Causality in Foundation World Models for Embodied AI
2105 - Toward Causal Representation Learning


2305 - Interventional Causal Representation Learning
2310 - Towards Causal Foundation Model: on Duality between Causal Inference and Attention
2203 - Weakly Supervised Causal Representation Learning
2202 - CITRIS: Causal Identifiability from Temporal Intervened Sequences
2100 - CausalVAE: Disentangled Representation Learning via Neural Structural Causal Models
2000 - iVAE: Variational Autoencoders and Nonlinear ICA: A Unifying Framework


2405 - Demystifying Amortized Causal Discovery with Transformers
2402 - Efficient Causal Graph Discovery Using Large Language Models
2204 - Learning to Induce Causal Structure

Causality in NeuroCog

2405 - The Development of Human Causal Learning and Reasoning
0000 - Beyond the here and now: Counterfactual Simulation in Causal Cognition (by Tobias Gerstenberg)

Causal World Model & Agent

2404 - Robust Agents Learn Causal World Models
2404 - Compete and Compose: Learning Independent Mechanisms for Modular World Models
2404 - Empowerment as Causal Learning, Causal Learning as Empowerment A bridge between Bayesian causal hypothesis testing and reinforcement learning
2403 - Dreaming of Many Worlds: Learning Contextual World Models Aids Zero-Shot Generalization
2200 - CausalDyna: Improving Generalization of Dyna-Style Reinforcement Learning via Counterfactual-Based Data Augmentation
2100 - Systematic Evaluation of Causal Discovery in Visual Model Based Reinforcement Learning

Planning & World Models


2310 - Skipper - Combining Spatial and Temporal Abstraction in Planning for Better Generalization [YB]

Position Papers

2405 - What is Lacking in Sora and V-JEPA's World Models? -A Philosophical Analysis of Video AIs Through the Theory of Productive Imagination
2403 - Are Video Generation Models World Simulators?

World Representations

2404 - V-JEPA - Revisiting Feature Prediction for Learning Visual Representations from Video
2307 - IVCL - Does Visual Pretraining Help End-to-End Reasoning?
2206 - BYOL-Explore: Exploration by Bootstrapped Prediction


2405 - Diffusion for World Modeling: Visual Details Matter in Atari
24xx - PlanDQ: Hierarchical Plan Orchestration via D-Conductor and Q-Performer
2403 - Subgoal Diffuser: Coarse-to-fine Subgoal Generation to Guide Model Predictive Control for Robot Manipulation
2402 - Stitching Sub-Trajectories with Conditional Diffusion Model for Goal-Conditioned Offline RL
2402 - DiffStitch: Boosting Offline Reinforcement Learning with Diffusion-based Trajectory Stitching
2401 - Simple Hierarchical Planning with Diffusion
2401 - Closing the Gap between TD Learning and Supervised Learning -- A Generalisation Point of View


2405 - CarDreamer: Open-source Learning Platform for World Model Based Autonomous Driving
2403 - Dreaming of Many Worlds: Learning Contextual World Models Aids Zero-Shot Generalization


2309 - Memory Gym: Towards Endless Tasks to Benchmark Memory Capabilities of Agents


2405 - CarDreamer: Open-source Learning Platform for World Model Based Autonomous Driving


2405 - AndroidWorld: A Dynamic Benchmarking Environment for Autonomous Agents
2402 - Craftax: A Lightning-Fast Benchmark for Open-Ended Reinforcement Learning



2405 - The Memory Systems of The Human Brain and Generative Artificial Intelligence
2405 - In Search of Dispersed Memories: Generative Diffusion Models Are Associative Memory Networks
2112 - Relating Transformers to Models and Neural Representations of The Hippocampal Formation [J. Whittington]


2206 - BYOL-Explore: Exploration by Bootstrapped Prediction

Safety & Alignment


2404 - Regulating Advanced Artificial Agents [Y. Bengio, Science]
24xx - Safeguarded AI: Constructing Guaranteed Safety
2310 - AI Alignment: A Comprehensive Survey [Peking University]


2404 - BIRD: A Trustworthy Bayesian Inference Framework for Large Language Models
2404 - Probabilistic Inference in Language Models via Twisted Sequential Monte Carlo
2404 - Your Language Model is Secretly a Q-Function
2312 - Training Chain-Of-Thought via Latent-Variable Inference
2310 - Amortizing Intractable Inference in Large Language Models [github]
2309 - Making Large Language Models Better Reasoners with Alignment
2304 - Bayesian Low-Rank Adaptation for Large Language Models [ICLR24]
2309 - Don’t Throw Away Your Value Model! Generating More Preferable Text with Value-Guided Monte-Carlo Tree Search Decoding
2205 - RL with KL Penalties is Better Viewed as Bayesian Inference
2202 - Red Teaming Language Models with Language Models
2111 - An Explanation of In-Context Learning as Implicit Bayesian Inference

Embodied Agent


2312 - Conformal Prediction for Uncertainty-Aware Planning with Diffusion Dynamics Model [NeurIPS’23]
2306 - Safe Planning with Diffusion Probabilistic Models

Safe Reinforcement Learning & Planning

2402 - Leveraging Approximate Model-based Shielding for Probabilistic Safety Guarantees in Continuous Environments
2307 - SafeDreamer: Safe Reinforcement Learning with World Models
2306 - Safe Planning with Diffusion Probabilistic Models
2306 - Trajectory Generation, Control, and Safety with Denoising Diffusion Probabilistic Models
2304 - Approximate Shielding of Atari Agents for Safe Exploration
2101 - Shielding Atari Games with Bounded Prescience

General Backbone Architectures

2405 - Aaren - Attention as an RNN
2212 - DiT - Scalable Diffusion Models with Transformers
2202 - GroupViT: Semantic Segmentation Emerges from Text Supervision
2202 - MaskGIT: Masked Generative Image Transformer


24XX - Maximum Entropy GFlowNets with Soft Q-Learning
24XX - Generative Flow Networks as Entropy-Regularized RL
2310 - Local Search GFlowNets
2209 - Learning GFlowNets from Partial Episodes for Improved Convergence and Stability

Protein Design

Position Papers

2402 - Generative AI for Controllable Protein Sequence Design: A Survey

LLM General

Position Papers

2402 - Position Paper: Bayesian Deep Learning in the Age of Large-Scale AI


2405 - Understanding Transformer Reasoning Capabilities via Graph Algorithms
2405 - From Explicit CoT to Implicit CoT: Learning to Internalize CoT Step by Step
2405 - Metacognitive Capabilities of LLMs- An Exploration in Mathematical Problem Solving
2404 - A Survey on Deep Learning for Theorem Proving
2403 - Machine Learning and Information Theory Concepts Towards an Ai Mathematician
2402 - Uncertainty of Thoughts: Uncertainty-Aware Planning Enhances Information Seeking in Large Language Models
2308 - Graph of Thoughts - Solving Elaborate Problems with Large Language Models
2212 - Prompting Is Programming- A Query Language for Large Language Models


2305 -   
2212 - Solving Quantitative Reasoning Problems with Language Models
2205 - Autoformalization with Large Language Models
20xx - A Promising Path Towards Autoformalization and General Artificial Intelligence

In Brain

1709 - Language, Mind and Brain