In this paper we study instruction following in text-based games and propose an approach that advances theprevious stateoftheart.

In Appendix A, we describe each of the components in GA T A in detail. In Appendix B, we provide detailed information on how we pre-train GA T A's graph updater GA T A. Since the action scorer module is the same as in GA T A, this appendix elaborates on In Appendix D, we provide additional results and discussions. In Appendix F, we show examples of graphs in TextWorld games. As briefly mentioned in Section 3.3, GA T A utilizes a graph encoder which is based on R-GCN [ The number of bases we use is 3. 14 A.2 T ext Encoder In the block, each convolutional layer has 64 filters, each kernel's size is 5. Following standard transformer training, we add positional encodings into each block's The representation aggregator aims to combine the text observation representations and graph representations together. The scorer consists of a self-attention layer, a masked mean pooling layer, and a two-layer MLP .

GATAis trained using acombination of reinforcement and self-supervised learning. Our workdemonstrates thatthelearned graph-based representations helpagents converge to better policies than their text-only counterparts and facilitate effective generalization across game configurations.

Belief systems are rarely globally consistent, yet effective reasoning often persists locally. We propose a novel graph-theoretic framework that cleanly separates credibility--external, a priori trust in sources--from confidence--an internal, emergent valuation induced by network structure. Beliefs are nodes in a directed, signed, weighted graph whose edges encode support and contradiction. Confidence is obtained by a contractive propagation process that mixes a stated prior with structure-aware influence and guarantees a unique, stable solution. Within this dynamics, we define reasoning zones: high-confidence, structurally balanced subgraphs on which classical inference is safe despite global contradictions. We provide a near-linear procedure that seeds zones by confidence, tests balance using a parity-based coloring, and applies a greedy, locality-preserving repair with Jaccard de-duplication to build a compact atlas. To model belief change, we introduce shock updates that locally downscale support and elevate targeted contradictions while preserving contractivity via a simple backtracking rule. Re-propagation yields localized reconfiguration-zones may shrink, split, or collapse--without destabilizing the entire graph. We outline an empirical protocol on synthetic signed graphs with planted zones, reporting zone recovery, stability under shocks, and runtime. The result is a principled foundation for contradiction-tolerant reasoning that activates classical logic precisely where structure supports it.

Playing text-based games requires skills in processing natural language and sequential decision making.

User prompts for generative AI models are often underspecified, leading to sub-optimal responses. This problem is particularly evident in text-to-image (T2I) generation, where users commonly struggle to articulate their precise intent. This disconnect between the user's vision and the model's interpretation often forces users to painstakingly and repeatedly refine their prompts. To address this, we propose a design for proactive T2I agents equipped with an interface to (1) actively ask clarification questions when uncertain, and (2) present their understanding of user intent as an understandable belief graph that a user can edit. We build simple prototypes for such agents and verify their effectiveness through both human studies and automated evaluation. We observed that at least 90% of human subjects found these agents and their belief graphs helpful for their T2I workflow. Moreover, we develop a scalable automated evaluation approach using two agents, one with a ground truth image and the other tries to ask as few questions as possible to align with the ground truth. On DesignBench, a benchmark we created for artists and designers, the COCO dataset (Lin et al., 2014), and ImageInWords (Garg et al., 2024), we observed that these T2I agents were able to ask informative questions and elicit crucial information to achieve successful alignment with at least 2 times higher VQAScore (Lin et al., 2024) than the standard single-turn T2I generation. Demo: https://github.com/google-deepmind/proactive_t2i_agents.

While large language models (LLMs) are proficient at question-answering (QA), it is not always clear how (or even if) an answer follows from their latent "beliefs". This lack of interpretability is a growing impediment to widespread use of LLMs. To address this, our goals are to make model beliefs and their inferential relationships explicit, and to resolve inconsistencies that may exist, so that answers are supported by interpretable chains of reasoning drawn from a consistent network of beliefs. Our approach, which we call REFLEX, is to add a rational, self-reflecting layer on top of the LLM. First, given a question, we construct a belief graph using a backward-chaining process to materialize relevant model beliefs (including beliefs about answer candidates) and their inferential relationships. Second, we identify and minimize contradictions in that graph using a formal constraint reasoner. We find that REFLEX significantly improves consistency (by 8%-11% absolute) without harming overall answer accuracy, resulting in answers supported by faithful chains of reasoning drawn from a more consistent belief system. This suggests a new style of system architecture in which an LLM extended with a rational layer can provide an interpretable window into system beliefs, add a systematic reasoning capability, and repair latent inconsistencies present in the LLM.

Robots often have to operate in discrete partially observable worlds, where the states of world are only observable at runtime. To react to different world states, robots need contingencies. However, computing contingencies is costly and often non-optimal. To address this problem, we develop the improved path tree optimization (PTO) method. PTO computes motion contingencies by constructing a tree of motion paths in belief space. This is achieved by constructing a graph of configurations, then adding observation edges to extend the graph to belief space. Afterwards, we use a dynamic programming step to extract the path tree. PTO extends prior work by adding a camera-based state sampler to improve the search for observation points. We also add support to non-euclidean state spaces, provide an implementation in the open motion planning library (OMPL), and evaluate PTO on four realistic scenarios with a virtual camera in up to 10-dimensional state spaces. We compare PTO with a default and with the new camera-based state sampler. The results indicate that the camera-based state sampler improves success rates in 3 out of 4 scenarios while having a significant lower memory footprint. This makes PTO an important contribution to advance the state-of-the-art for discrete belief space planning.