Beyond the Current Observation: Evaluating Multimodal Large Language Models in Controllable Non-Markov Games

• URL: http://arxiv.org/abs/2606.19338v1
• Date: Wed, 17 Jun 2026 17:59:34 GMT
• Title: Beyond the Current Observation: Evaluating Multimodal Large Language Models in Controllable Non-Markov Games
• Abstract summary: We introduce RNG-Bench, a benchmark suite designed to isolate a base model's ability to reconstruct past observations.<n>RNG-Bench includes two complementary games: Matching Pairs, where card identities briefly revealed at specific locations must later be recalled, and 3D Maze, where egocentric views must be integrated into a spatial map.<n>The hardest configurations require contexts of roughly 128K tokens and 350 image inputs per episode, and remain far from saturated by frontier MLLMs.
• Score: 69.57330692969543
• License: http://creativecommons.org/licenses/by-nc-sa/4.0/
• Abstract: Deploying multimodal foundation models as closed-loop policies increasingly requires conditioning actions on observations that are no longer visible. However, existing benchmarks either expose the full state, conflate hidden-state reconstruction with other agent skills, or test recall only after an episode has ended. We introduce RNG-Bench (Reconstructive Non-Markov Games), a benchmark suite designed to isolate a base model's ability to reconstruct past observations and act on them during multi-step interaction. RNG-Bench includes two complementary games: Matching Pairs, where card identities briefly revealed at specific locations must later be recalled, and 3D Maze, where egocentric views must be integrated into a spatial map. Both games are evaluated under a unified harness with three controlled difficulty axes: grid size, visual pattern, and observation modality. The benchmark further introduces a head-to-head duel protocol to control for instance-level variance and a Memory Gap metric that disentangles forgetting from poor action selection. The hardest configurations require contexts of roughly 128K tokens and 350 image inputs per episode, and remain far from saturated by frontier MLLMs. Memory Gap analysis shows that most residual errors stem from forgetting earlier observations rather than from suboptimal decision making. Finally, fine-tuning Qwen3.5-9B on optimal-policy rollouts and filtered model demonstrations improves performance on RNG-Bench and transfers to existing benchmarks without degrading general multimodal capability.

Related papers

• MNIST-PRO: MNIST is Back as a Partially Observable World for AI Agents [74.83535434786145]
  MNIST-PRO is a benchmark that isolates agentic perception by converting MNIST digit recognition into a sequential, glimpse-based search task with lookback constraints.<n>We evaluate ten multimodal models across four memory representations, including raw visual history, textual states, structured metric grid maps, and a consolidated visual canvas.
  arXiv  Detail & Related papers  (2026-08-31T16:05:39Z)
• GamiBench: Evaluating Spatial Reasoning and 2D-to-3D Planning Capabilities of MLLMs with Origami Folding Tasks [2.037026133182214]
  We introduce GamiBench, a benchmark to evaluate spatial reasoning in large language models (MLLMs) through origami-inspired folding tasks.<n>GamiBench includes 186 regular and impossible 2D crease patterns paired with their corresponding 3D folded shapes.<n>GamiBench holistically evaluates the entire reasoning process--measuring cross-view consistency, physical feasibility through impossible-fold detection, and interpretation of intermediate folding steps.
  arXiv  Detail & Related papers  (2025-12-22T01:07:59Z)
• Detect Anything via Next Point Prediction [51.55967987350882]
  Rex- Omni is a 3B-scale MLLM that achieves state-of-the-art object perception performance.<n>On benchmarks like COCO and LVIS, Rex- Omni attains performance comparable to or exceeding regression-based models.
  arXiv  Detail & Related papers  (2025-10-14T17:59:54Z)
• MR$^2$-Bench: Going Beyond Matching to Reasoning in Multimodal Retrieval [86.35779264575154]
  Multimodal retrieval is becoming a crucial component of modern AI applications, yet its evaluation lags behind the demands of more realistic and challenging scenarios.<n>We introduce MR$2$-Bench, a reasoning-intensive benchmark for multimodal retrieval.
  arXiv  Detail & Related papers  (2025-09-30T15:09:14Z)
• OST-Bench: Evaluating the Capabilities of MLLMs in Online Spatio-temporal Scene Understanding [50.72259772580637]
  We introduce OST-Bench, a benchmark designed to evaluate Online Spatio-Temporal understanding from the perspective of an agent.<n>Built on an efficient data collection pipeline, OST-Bench consists of 1.4k scenes and 10k question-answer pairs collected from ScanNet, Matterport3D, and ARKitScenes.<n>We find that both complex-based spatial reasoning demands and long-term memory retrieval requirements significantly drop model performance along two separate axes.
  arXiv  Detail & Related papers  (2025-07-10T17:56:07Z)
• TurnBench-MS: A Benchmark for Evaluating Multi-Turn, Multi-Step Reasoning in Large Language Models [5.6525926183880255]
  We introduce TurnBench, a novel benchmark that evaluates multi-turn, multi-step reasoning through an interactive code-breaking task.<n>In each episode, a model must uncover hidden logical or arithmetic rules by making sequential guesses, receiving structured feedback, and integrating clues across multiple rounds.<n>TurnBench includes two modes: Classic, which tests standard reasoning, and Nightmare, which introduces increased complexity and requires robust inferential chains.
  arXiv  Detail & Related papers  (2025-06-02T05:47:50Z)
• Model as a Game: On Numerical and Spatial Consistency for Generative Games [117.36098212829766]
  We revisit the paradigm of generative games to explore what truly constitutes a Model as a Game (MaaG) with a well-developed mechanism.<n>Based on the DiT architecture, we design two specialized modules: (1) a numerical module that integrates a LogicNet to determine event triggers, with calculations processed externally as conditions for image generation; and (2) a spatial module that maintains a map of explored areas, retrieving location-specific information during generation and linking new observations to ensure continuity.
  arXiv  Detail & Related papers  (2025-03-27T05:46:15Z)
• When Liebig's Barrel Meets Facial Landmark Detection: A Practical Model [87.25037167380522]
  We propose a model that is accurate, robust, efficient, generalizable, and end-to-end trainable. In order to achieve a better accuracy, we propose two lightweight modules. DQInit dynamically initializes the queries of decoder from the inputs, enabling the model to achieve as good accuracy as the ones with multiple decoder layers. QAMem is designed to enhance the discriminative ability of queries on low-resolution feature maps by assigning separate memory values to each query rather than a shared one.
  arXiv  Detail & Related papers  (2021-05-27T13:51:42Z)

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.