Pith. sign in

REVIEW 2 major objections 60 references

A new benchmark reveals that current interactive world models lack reliable long-horizon stability across multiple dimensions.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

T0 review · grok-4.3

2026-07-03 21:56 UTC pith:KWMJVSH3

load-bearing objection The paper defines four new metric families for long-horizon world model stability but supplies no correlation checks or representativeness stats to back them up. the 2 major comments →

arxiv 2606.31672 v2 pith:KWMJVSH3 submitted 2026-06-30 cs.CV cs.AI

WorldOdysseyBench: An Open-World Benchmark for Long-Horizon Stability of Interactive World Models

Ting-Bing Xu , Jiacheng Sui , Zhe Gao , Kewei Shi , Wenjin Yang , Zhicheng Liu , Zhaoxu Sun , Mingchao Sun
show 7 more authors
Hongyu Pan Fan Jiang Mu Xu Qi Fan Yang Gao Yong Li Baoquan Chen
This is my paper
classification cs.CV cs.AI
keywords interactive world modelslong-horizon stabilitybenchmarkaction evaluationphysics consistencymemory assessmentopen-world scenesvisual drift
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper presents WorldOdysseyBench to evaluate interactive world models on sustained performance over extended interactions rather than short action sequences. It introduces four metrics that separately track per-frame actions, visual changes within segments, physics behavior only when actions are followed correctly, and memory of scenes and objects without relying on ongoing actions. Tests on more than ten models across hundreds of cases in varied environments show that none maintain consistency in all areas at once. This evaluation matters because models that cannot hold stable representations over time cannot support applications needing continuous interaction with changing environments.

Core claim

Existing benchmarks evaluate action following only at the trajectory level while ignoring memory and interaction physics. WorldOdysseyBench addresses this with four dimensions: per-frame action metric to expose hidden failures, segment-based drift for non-monotonic collapse, controllability-gated physics for plausibility, and action-decoupled memory via 3D reconstruction and tracking. Evaluation of over ten models finds none reliably satisfies all, with the best scoring only moderately on the 600+ test cases in nature, urban, and indoor environments.

What carries the argument

WorldOdysseyBench benchmark with its four specialized metrics for per-frame action accuracy, segment-based visual drift, controllability-gated physics plausibility, and action-decoupled memory assessment through 3D point clouds and tracking.

Load-bearing premise

The four metrics correctly measure the intended aspects of long-horizon stability and the test cases represent open-world interactions.

What would settle it

A model that scores high across all four metrics yet fails in extended real interactive tasks, or a model that fails the benchmark metrics but succeeds in sustained real-world use.

Watch this falsifier — get emailed when new claim-graph text bears on it.

If this is right

  • Models that improve on the four metrics would support more reliable deployment in applications requiring continuous interaction.
  • Trajectory-level scores can conceal per-frame action mistakes and sudden visual collapses in the middle of sequences.
  • Physics checks must be gated on accurate action execution to isolate true physical consistency from control errors.
  • Memory tests performed without action cues can separately verify recall of scenes and tracked subjects.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The benchmark design could push development toward models that combine learned prediction with separate memory buffers.
  • Longer test sequences beyond 60 seconds might expose additional memory decay patterns not visible in the current cases.
  • Direct transfer of high-scoring models to physical robot control could test whether benchmark results predict real stability.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit.

Referee Report

2 major / 0 minor

Summary. The paper introduces WorldOdysseyBench, an open-world benchmark for long-horizon stability of interactive world models (IWMs). It defines four tailored metrics—per-frame action (bypassing semantic scale issues), segment-based drift for vision (capturing mid-sequence collapse), controllability-gated physics (over mechanics/optics/3D), and action-decoupled memory (via 3D reconstruction and tracking+VLM)—evaluated on 600+ test cases spanning Nature/Urban/Indoor scenes in first/third-person views with 10-60s WASD interactions. Evaluation of 10+ open/closed-source models finds none reliably satisfy all dimensions, with the best achieving only moderate scores.

Significance. If the metrics prove valid and the test cases representative, the benchmark would be a useful contribution by exposing stability, physics, and memory failures missed by prior trajectory-level or start-vs-end evaluations, providing a concrete framework to drive progress toward deployable IWMs.

major comments (2)
  1. [Abstract] Abstract: the central claim that the four metrics 'capture failures missed by prior trajectory-level or start-vs-end metrics' and that 'none reliably satisfies all dimensions' is load-bearing on the unverified assumption that per-frame action, segment-based drift, controllability-gated physics, and action-decoupled memory correctly quantify the intended aspects; no correlation study, ablation against existing metrics, or human-judgment validation is referenced.
  2. [Abstract] Abstract: the representativeness of the 600+ test cases for open-world interaction is asserted without supporting diversity statistics (e.g., scene-type distribution, duration histogram, viewpoint balance), which directly affects whether the headline result generalizes beyond the chosen cases.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive feedback. We address each major comment below and will revise the manuscript accordingly to strengthen the presentation of the benchmark.

read point-by-point responses

  1. Referee: [Abstract] Abstract: the central claim that the four metrics 'capture failures missed by prior trajectory-level or start-vs-end metrics' and that 'none reliably satisfies all dimensions' is load-bearing on the unverified assumption that per-frame action, segment-based drift, controllability-gated physics, and action-decoupled memory correctly quantify the intended aspects; no correlation study, ablation against existing metrics, or human-judgment validation is referenced.

    Authors: The metrics were designed to target concrete, observable limitations of prior evaluations, as motivated in the introduction: per-frame action avoids semantic scale disparities that affect trajectory-level scores, segment-based drift detects mid-sequence non-monotonic collapses invisible to start-vs-end comparisons, controllability-gated physics isolates physical plausibility under correct action execution, and action-decoupled memory separates scene and subject recall from action following. The empirical results across 10+ models illustrate distinct failure patterns across these axes that standard metrics do not surface. We agree, however, that explicit supporting analyses were not included in the submission. In the revision we will add an appendix containing (i) an ablation comparing the new metrics against trajectory-level action accuracy, FVD, and start-vs-end drift on the same model outputs and (ii) a small-scale human preference study on a 50-case subset to check alignment with perceived stability, physics, and memory errors. revision: yes

  2. Referee: [Abstract] Abstract: the representativeness of the 600+ test cases for open-world interaction is asserted without supporting diversity statistics (e.g., scene-type distribution, duration histogram, viewpoint balance), which directly affects whether the headline result generalizes beyond the chosen cases.

    Authors: We agree that quantitative diversity statistics would make the claim of representativeness more transparent. The test cases were collected to span Nature/Urban/Indoor environments, first- and third-person viewpoints, and 10–60 s durations, but these were described qualitatively rather than with explicit distributions. In the revised manuscript we will add a dedicated subsection (or table) in the benchmark description reporting scene-type counts, a duration histogram, and first/third-person balance across the full set of 600+ cases. revision: yes

Circularity Check

0 steps flagged

No significant circularity; benchmark definition paper

full rationale

This is a benchmark construction paper that defines four new evaluation metrics (per-frame action, segment-based drift, controllability-gated physics, action-decoupled memory) and applies them to existing models. No equations, parameter fits, or predictions are present that could reduce to the inputs by construction. The central claims rest on the explicit definitions of the metrics and the collection of 600+ test cases rather than any self-referential derivation or load-bearing self-citation chain. The evaluation results follow directly from applying the stated protocols and do not presuppose the outcomes.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Abstract-only; no explicit free parameters, axioms, or invented entities are stated. The benchmark implicitly assumes the chosen scenes and interaction lengths are sufficient proxies for open-world stability.

pith-pipeline@v0.9.1-grok · 5803 in / 1007 out tokens · 23184 ms · 2026-07-03T21:56:48.608659+00:00 · methodology

0 comments
read the original abstract

Despite rapid progress in interactive world models (IWMs), existing benchmarks evaluate action following only at trajectory level and ignore memory and interaction physics. We introduce WorldOdysseyBench, an open-world benchmark for long-horizon stability across four dimensions, each with tailored innovations: (i) Action: per-frame action metric bypassing cross-model semantic scale disparity and exposing failures hidden by trajectory; (ii) Vision: segment-based drift metric capturing non-monotonic mid-sequence collapse missed by start-vs-end comparisons; (iii) Physics: controllability-gated evaluation over mechanics, optics, and 3D consistency, scoring plausibility under faithful action execution; (iv) Memory: action-decoupled protocol evaluating scene memory via transition-localized 3D point-cloud reconstruction and subject memory via tracking-plus-VLM reasoning. The benchmark comprises 600+ test cases across Nature, Urban, and Indoor scenes in first/third-person views with WASD 10-60s continuous interaction. Evaluating 10+ open/closed-source models reveals none reliably satisfies all dimensions; even the best achieves only moderate scores. Advances on WorldOdysseyBench are steps toward IWMs that are stable, physically grounded, memory-faithful, and deployable in real-world applications.

Figures

Figures reproduced from arXiv: 2606.31672 by Baoquan Chen, Fan Jiang, Hongyu Pan, Jiacheng Sui, Kewei Shi, Mingchao Sun, Mu Xu, Qi Fan, Ting-Bing Xu, Wenjin Yang, Yang Gao, Yong Li, Zhaoxu Sun, Zhe Gao, Zhicheng Liu.

Figure 1
Figure 1. Figure 1: Illustration of WorldRoamBench. WorldRoamBench evaluates interactive world models across action following, visual quality, memory, and interaction physics, covering open-source and closed-source models in diverse scenarios and viewing conditions. Abstract Despite rapid progress in interactive world models (IWMs), existing benchmarks evaluate action following only at tra￾jectory level and ignore memory and … view at source ↗
Figure 2
Figure 2. Figure 2: Failures revealed by long-horizon interaction. Ex￾tended rollouts expose failures often missed by short clips: high trajectory scores hide per-step action mismatches, visual quality degrades, physical constraints are violated, and revisited scenes are regenerated inconsistently. Yet as these models proliferate, a critical question re￾mains unanswered: How well do they respond to user inputs over extended i… view at source ↗
Figure 3
Figure 3. Figure 3: Overview of the WorldRoamBench evaluation pipeline. Given videos generated from shared initial frames and WASD/IJKL action programs, WorldRoamBench evaluates four complementary dimensions: action following through pose-aligned frame scoring, visual quality through frame-level and drift-aware metrics, interaction physics through mechanics, optics, and 3D-consistency tests, and memory through turning-point-a… view at source ↗
Figure 4
Figure 4. Figure 4: Action following evaluation pipeline. A shared ViPE trajectory estimation stage feeds two branches: (left) per-frame ac￾tion accuracy via latent-stride discretization, and (right) TrajScore via adaptive GT construction and arc-length resampling. an atomic action (e.g., forward) or a compound action. Each action is decomposed into a set of atomic sub-actions P(gt) = {p1, p2, . . .}. The atomic vocabulary co… view at source ↗
Figure 5
Figure 5. Figure 5: Memory evaluation pipeline. WorldRoamBench eval￾uates memory with two trajectory-aware tracks. Scene memory localizes the executed observation–revisit transition and compares reconstructed scene geometry across the two segments. Subject memory tracks the third-person protagonist and evaluates identity, structure, and appearance preservation with a holistic Qwen3-VL￾PLUS judgment. The full pipeline is provi… view at source ↗
Figure 6
Figure 6. Figure 6: Action key distribution across the three evaluation di￾mensions. Each pie chart shows the proportion and count of per￾frame key presses aggregated over all test cases in that dimension. 0 50 100 150 200 Number of Test Cases Memory Action Physics 99 113 212 118 99 217 96 75 171 (a) Case Counts FPV TPV 0 300 300 400 400 500 500 1000 1000 1100 0 100 200 300 400 Number of Cases 115 349 107 19 10 (b) Action Seq… view at source ↗
Figure 7
Figure 7. Figure 7: Test case count and action-sequence length distribu￾tion. (a) Test cases by dimension, split by 1st and 3rd perspective. (b) Distribution of action-sequence lengths in frames. related artifacts. 4.3. Test Suite Statistics [PITH_FULL_IMAGE:figures/full_fig_p011_7.png] view at source ↗
Figure 9
Figure 9. Figure 9: Per-frame visual quality curves and drift comparison (first-person). Top: mean imaging and aesthetic scores at each frame index (averaged over 120 videos per model, first 300 frames). Bottom: drift scores quantifying quality degradation from the best to the worst segment of the rollout (lower = more stable); bar height is the mean drift across videos, and black error bars span from mean−1 std (lower cap) t… view at source ↗
Figure 10
Figure 10. Figure 10: Per-frame action accuracy by difficulty (first-person). Strict accuracy (left) and partial accuracy (right) grouped by difficulty level (easy = constant action, medium = 1 action switch, hard = 2 action switches). Bar height is the mean across test cases. Most models degrade from easy to hard, but the magnitude of degradation varies substantially. stantially stronger third-person control than the open mod… view at source ↗
Figure 11
Figure 11. Figure 11: Test suite gallery. Each panel shows a representative test case with the first frame, trajectory schematic, and action sequence. The gallery is organized by scene category from top to bottom (Indoor, Urban, Nature) and by perspective from left to right (first-person, third-person), yielding six panels. checkpoint, which generates frames conditioned on both past and future context within each chunk, and ev… view at source ↗
Figure 12
Figure 12. Figure 12: Closed-source automated interaction pipeline for Happy Oyster and Genie 3. The system reads benchmark test cases (action [PITH_FULL_IMAGE:figures/full_fig_p021_12.png] view at source ↗
Figure 13
Figure 13. Figure 13: Latent-stride single-frame discretization. [PITH_FULL_IMAGE:figures/full_fig_p022_13.png] view at source ↗
Figure 15
Figure 15. Figure 15: Detailed interaction-physics evaluation pipeline. The full pipeline includes action-responsiveness validation, me￾chanics protocols, optics protocols, 3D-consistency evaluation, Qwen3-VL-PLUS-based VLM scoring, fallback queries, and domain-level aggregation. execute the prescribed action rather than remaining static or unresponsive. This prevents action-following failures from being conflated with physica… view at source ↗
Figure 16
Figure 16. Figure 16: Detailed memory evaluation pipeline. The full pipeline includes action-aware transition localization, point-cloud reconstruc￾tion, registration, filtering, scene-memory scoring, subject tracking, controllability gating, and Qwen3-VL-PLUS-based VLM subject￾memory evaluation. camera-pose chain. Pose drift can introduce a rigid dis￾placement unrelated to memory, artificially increasing both forgetting and ha… view at source ↗
Figure 17
Figure 17. Figure 17: Failure mode of frame-pair memory evaluation. Frame-pair metrics assume that an observation frame and its re￾visit counterpart depict the same spatial location. In long-horizon interactive rollouts, however, imperfect action execution can shift the revisit trajectory, so the paired frames correspond to different viewpoints or even different scene regions. Image-level discrep￾ancies in such pairs therefore… view at source ↗
Figure 18
Figure 18. Figure 18: Action gap in closed-source models. Top: Genie 3. When the D key (rightward translation) is issued, the model rotates the camera to the right instead of translating the character rightward within a stable scene. The character remains centered while the entire scene rotates, making controlled navigation impossible. Bottom: Happy Oyster. When the D key (rightward translation) is issued, the model faithfully… view at source ↗
Figure 19
Figure 19. Figure 19: Qualitative examples of WorldRoamBench scoring across physics, memory, and action. Each block contrasts a positive and a negative rollout that share the same initial frame and action schedule. Physics: W+A in a corridor—Happy Oyster (No Clipping) vs. Genie 3 (Clipping). Memory: J→L Observation/Revisit in an alley—Matrix-Game 3.0 (Good Memory) vs. SANA-WM (Bad Memory). Action: S→A (backward→left)—Genie 3 (… view at source ↗
Figure 20
Figure 20. Figure 20: Collision evaluation prompts. Full VLM prompts used for approach detection and collision-response verification. Deformation Evaluation Prompt — FPV: Trace Detection You are analyzing a video clip ({duration}s total). Showing {N} frames at timestamps: [{t1, t2, ..., tN}]. Think step by step about the following question. Question: In these frames from the later part of a first-person walking video, look car… view at source ↗
Figure 21
Figure 21. Figure 21: Deformation evaluation prompts. Full VLM prompts used for first-person trace detection, third-person reference comparison, and third-person real-time interaction checks. 35 [PITH_FULL_IMAGE:figures/full_fig_p035_21.png] view at source ↗
Figure 22
Figure 22. Figure 22: Clipping evaluation prompts. Full VLM prompts used by the staged clipping-detection cascade. 36 [PITH_FULL_IMAGE:figures/full_fig_p036_22.png] view at source ↗
Figure 23
Figure 23. Figure 23: Gravity evaluation prompts. Full VLM prompts used for scenario-specific gravity checks and the third-person general gravity check. 37 [PITH_FULL_IMAGE:figures/full_fig_p037_23.png] view at source ↗
Figure 24
Figure 24. Figure 24: Terrain-following evaluation prompts. Full VLM prompts used for naturalness, third-person ground contact, and boundary checks. 38 [PITH_FULL_IMAGE:figures/full_fig_p038_24.png] view at source ↗
Figure 25
Figure 25. Figure 25: Reflection evaluation prompts. Full VLM prompts used for coarse reflection plausibility checks and per-frame reflection scoring. 39 [PITH_FULL_IMAGE:figures/full_fig_p039_25.png] view at source ↗
Figure 26
Figure 26. Figure 26: Occlusion and shadow evaluation prompts. Full VLM prompts used for expected-shadow, fallback shadow, and generic shadow checks. 40 [PITH_FULL_IMAGE:figures/full_fig_p040_26.png] view at source ↗
Figure 27
Figure 27. Figure 27: Subject memory evaluation prompt. Full VLM prompt used for holistic third-person subject-memory scoring and diagnostic flag extraction. 41 [PITH_FULL_IMAGE:figures/full_fig_p041_27.png] view at source ↗

Review history (2 revisions) →

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

60 extracted references · 60 canonical work pages · 7 internal anchors

  1. [1]

    Genie: Generative interactive environments

    Bruce, J., Dennis, M., Edwards, A., et al. Genie: Generative interactive environments. InICML, 2024

    work page 2024

  2. [2]

    Genie 3: A new frontier for world models.https://deepmind.google/discover/ blog/genie- 3- a- new- frontier- for- world- models/, 2025

    Google DeepMind. Genie 3: A new frontier for world models.https://deepmind.google/discover/ blog/genie- 3- a- new- frontier- for- world- models/, 2025

    work page 2025

  3. [3]

    Happy Oyster: An open-ended world model for real-time world creation and interaction.https:// happyoyster.cn/, 2026

    Alibaba Group. Happy Oyster: An open-ended world model for real-time world creation and interaction.https:// happyoyster.cn/, 2026

    work page 2026

  4. [4]

    Wan: Open and Advanced Large-Scale Video Generative Models

    Wan Team. Wan 2.1: A comprehensive and unified video generation model.arXiv preprint arXiv:2503.20314, 2025

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  5. [5]

    Kling 3.0: Next-generation AI video generation

    Kuaishou. Kling 3.0: Next-generation AI video generation. https://klingai.com/, 2025

    work page 2025

  6. [6]

    Sora 2: A large-scale video generation model

    OpenAI. Sora 2: A large-scale video generation model. https://openai.com/sora/, 2025

    work page 2025

  7. [7]

    Veo 3: State-of-the-art video generation

    Google DeepMind. Veo 3: State-of-the-art video generation. https : / / deepmind . google / technologies / veo/, 2025

    work page 2025

  8. [8]

    Monkeyocr: Document parsing with a structure-recognition-relation triplet paradigm, 2026

    ByteDance Seed Team. Seedance 2.0: Advancing video generation for world complexity.arXiv preprint arXiv:2506.05218, 2025

    work page arXiv 2025

  9. [9]

    Matrix-game 2.0: An open-source real-time and streaming interactive world model

    Matrix-Game Team. Matrix-Game 2.0: An open-source, real-time, and streaming interactive world model.arXiv preprint arXiv:2508.13009, 2025

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  10. [10]

    Matrix-Game 3.0: Real-time and streaming interactive world model with long-horizon mem- ory.arXiv preprint, 2026

    Matrix-Game Team. Matrix-Game 3.0: Real-time and streaming interactive world model with long-horizon mem- ory.arXiv preprint, 2026

    work page 2026

  11. [11]

    HY-World 1.5: A systematic framework for interactive world modeling with real-time latency and ge- ometric consistency.arXiv preprint, 2025

    HY-World Team. HY-World 1.5: A systematic framework for interactive world modeling with real-time latency and ge- ometric consistency.arXiv preprint, 2025

    work page 2025

  12. [12]

    Yume 1.5: A text-controlled interactive world generation model.arXiv preprint, 2026

    Yume Team. Yume 1.5: A text-controlled interactive world generation model.arXiv preprint, 2026

    work page 2026

  13. [13]

    Advancing open-source world models.arXiv preprint, 2026

    LingBot Team. Advancing open-source world models.arXiv preprint, 2026. 16

    work page 2026

  14. [14]

    arXiv preprint arXiv:2602.08025 (2026)

    Ye, H., Lu, J., et al. MIND: Benchmarking memory consis- tency and action following in world models.arXiv preprint arXiv:2602.08025, 2026

    work page arXiv 2026

  15. [15]

    WorldMark: A Unified Benchmark Suite for Interactive Video World Models

    Alaya Studio. WorldMark: A unified benchmark suite for interactive video world models.arXiv preprint arXiv:2604.21686, 2026

    work page internal anchor Pith review Pith/arXiv arXiv 2026

  16. [16]

    iWorld-Bench: A benchmark for interactive world models with a unified action generation framework

    Li, Y ., et al. iWorld-Bench: A benchmark for interactive world models with a unified action generation framework. InICML, 2026

    work page 2026

  17. [17]

    Wildworld: A large-scale dataset for dynamic world modeling with actions and explicit state toward generative ARPG.arXiv preprint arXiv:2603.23497, 2026

    Shanda AI. WildWorld: A large-scale dataset for dynamic world modeling with actions and explicit state toward gener- ative ARPG.arXiv preprint arXiv:2603.23497, 2026

    work page arXiv 2026

  18. [18]

    VBench: Comprehensive benchmark suite for video generative mod- els

    Huang, Z., He, Y ., Yu, J., Zhang, F., Si, C., et al. VBench: Comprehensive benchmark suite for video generative mod- els. InCVPR, 2024

    work page 2024

  19. [19]

    VBench++: Comprehensive and versatile benchmark suite for video generative models.arXiv preprint arXiv:2411.13503, 2024

    Huang, Z., Zhang, F., Xu, X., He, Y ., Yu, J., et al. VBench++: Comprehensive and versatile benchmark suite for video gen- erative models.arXiv preprint arXiv:2411.13503, 2024

    work page arXiv 2024

  20. [20]

    VBench-2.0: Advancing Video Generation Benchmark Suite for Intrinsic Faithfulness

    Zheng, D., Huang, Z., Liu, H., Zou, K., He, Y ., et al. VBench-2.0: Advancing video generation benchmark suite for intrinsic faithfulness.arXiv preprint arXiv:2503.21755, 2025

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  21. [21]

    WorldScore: A unified evaluation benchmark for world generation

    Stanford. WorldScore: A unified evaluation benchmark for world generation. InICCV, 2025

    work page 2025

  22. [22]

    WorldModelBench: Judging video generation models as world models

    UC Berkeley. WorldModelBench: Judging video generation models as world models. InNeurIPS, 2025

    work page 2025

  23. [23]

    Towards world simulator: Crafting physical commonsense-based benchmark for video generation

    SJTU, et al. Towards world simulator: Crafting physical commonsense-based benchmark for video generation. In ICML, 2025

    work page 2025

  24. [24]

    WorldBench: Disambiguating physics for di- agnostic evaluation of world models.arXiv preprint arXiv:2601.21282, 2026

    UCLA. WorldBench: Disambiguating physics for di- agnostic evaluation of world models.arXiv preprint arXiv:2601.21282, 2026

    work page arXiv 2026

  25. [25]

    Video PreTraining (VPT): Learning to act by watching unlabeled online videos

    Baker, B., et al. Video PreTraining (VPT): Learning to act by watching unlabeled online videos. InNeurIPS, 2022

    work page 2022

  26. [26]

    Mineworld: A real-time and open-source interactive world model on Minecraft.arXiv preprint arXiv:2504.08388, 2025

    Microsoft. MineWorld: A real-time and open-source interactive world model on Minecraft.arXiv preprint arXiv:2504.08388, 2025

    work page arXiv 2025

  27. [27]

    SANA-WM: Efficient minute-scale world model- ing with hybrid linear diffusion transformer.arXiv preprint, 2026

    NVIDIA. SANA-WM: Efficient minute-scale world model- ing with hybrid linear diffusion transformer.arXiv preprint, 2026

    work page 2026

  28. [28]

    Lyra 2.0: Explorable generative 3D worlds.arXiv preprint, 2026

    NVIDIA. Lyra 2.0: Explorable generative 3D worlds.arXiv preprint, 2026

    work page 2026

  29. [29]

    minWM: A full-stack open-source frame- work for real-time interactive video world models.arXiv preprint, 2026

    minWM Team. minWM: A full-stack open-source frame- work for real-time interactive video world models.arXiv preprint, 2026

    work page 2026

  30. [30]

    LAION- 5B: An open large-scale dataset for training next generation image-text models

    Schuhmann, C., Beaumont, R., Vencu, R., et al. LAION- 5B: An open large-scale dataset for training next generation image-text models. InNeurIPS, 2022

    work page 2022

  31. [31]

    MUSIQ: Multi-scale image quality transformer

    Ke, J., Wang, Q., Wang, Y ., Milanfar, P., and Yang, F. MUSIQ: Multi-scale image quality transformer. InICCV, 2021

    work page 2021

  32. [32]

    Helios: A comprehensive benchmark for video generative models.arXiv preprint, 2025

    Helios Team. Helios: A comprehensive benchmark for video generative models.arXiv preprint, 2025

    work page 2025

  33. [33]

    WorldCompass: Reinforcement learning for long-horizon world models.arXiv preprint, 2026

    WorldCompass Team. WorldCompass: Reinforcement learning for long-horizon world models.arXiv preprint, 2026

    work page 2026

  34. [34]

    ViPE: Visual pose estimation for camera trajectory recovery

    work page

  35. [35]

    and Deng, J

    Teed, Z. and Deng, J. DROID-SLAM: Deep visual SLAM for monocular, stereo, and RGB-D cameras. InNeurIPS, 2021

    work page 2021

  36. [36]

    WBench: A comprehensive benchmark for evaluating world models via action-conditioned video gener- ation.arXiv preprint, 2026

    Ying, Z., et al. WBench: A comprehensive benchmark for evaluating world models via action-conditioned video gener- ation.arXiv preprint, 2026

    work page 2026

  37. [37]

    and Medioni, G

    Chen, Y . and Medioni, G. Object modelling by registra- tion of multiple range images.Image and Vision Computing, 10(3):145–155, 1992

    work page 1992

  38. [38]

    B., Blodow, N., and Beetz, M

    Rusu, R. B., Blodow, N., and Beetz, M. Fast Point Feature Histograms (FPFH) for 3D registration. InICRA, 2009

    work page 2009

  39. [39]

    Point Transformer V3: Simpler, faster, stronger

    Wu, X., Jiang, L., Wang, P.-S., et al. Point Transformer V3: Simpler, faster, stronger. InCVPR, 2024

    work page 2024

  40. [40]

    SAM 2: Segment Anything in Images and Videos

    Ravi, N., Gabeur, V ., Hu, Y .-T., et al. SAM 2: Seg- ment anything in images and videos.arXiv preprint arXiv:2408.00714, 2024

    work page internal anchor Pith review Pith/arXiv arXiv 2024

  41. [41]

    Two-frame motion estimation based on poly- nomial expansion

    Farneb ¨ack, G. Two-frame motion estimation based on poly- nomial expansion. InScandinavian Conference on Image Analysis (SCIA), 2003

    work page 2003

  42. [42]

    What limits virtual agent appli- cation? OmniBench: A scalable multi-dimensional bench- mark for essential virtual agent capabilities

    Bu, W., Wu, Y ., Yu, Q., et al. What limits virtual agent appli- cation? OmniBench: A scalable multi-dimensional bench- mark for essential virtual agent capabilities. InICML, 2025

    work page 2025

  43. [43]

    Omni-WorldBench: Towards a comprehensive interaction-centric evaluation for world models.arXiv preprint arXiv:2603.22212, 2026

    Wu, M., Cai, Z., Zhao, F., et al. Omni-WorldBench: Towards a comprehensive interaction-centric evaluation for world models.arXiv preprint arXiv:2603.22212, 2026

    work page arXiv 2026

  44. [44]

    Do vision-language models have internal world models? Towards an atomic evaluation

    Gao, Q., Pi, X., Liu, K., et al. Do vision-language models have internal world models? Towards an atomic evaluation. InACL, 2025

    work page 2025

  45. [45]

    How far is video generation from world model: A physical law perspective

    Kang, B., Yue, Y ., Lu, R., et al. How far is video generation from world model: A physical law perspective. InICML, 2025

    work page 2025

  46. [46]

    WorldLens: Full-Spectrum Evaluations of Driving World Models in Real World

    Liang, A., Kong, L., Yan, T., et al. WorldLens: Full-spectrum evaluations of driving world models in real world.arXiv preprint arXiv:2512.10958, 2025

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  47. [47]

    Worldarena: A unified benchmark for evaluating perception and functional utility of embodied world models.arXiv preprint arXiv:2602.08971, 2026

    Shang, Y ., Li, Z., Ma, Y ., et al. WorldArena: A unified benchmark for evaluating perception and functional utility of embodied world models.arXiv preprint arXiv:2602.08971, 2026

    work page arXiv 2026

  48. [48]

    ACT-Bench: Towards action controllable world models for autonomous driving.arXiv preprint arXiv:2412.05337, 2024

    Arai, H., Ishihara, K., Takahashi, T., and Yamaguchi, Y . ACT-Bench: Towards action controllable world models for autonomous driving.arXiv preprint arXiv:2412.05337, 2024

    work page arXiv 2024

  49. [49]

    EWMBench: Evaluat- ing scene, motion, and semantic quality in embodied world models

    Hu, Y ., Huang, S., Liao, Y ., et al. EWMBench: Evaluat- ing scene, motion, and semantic quality in embodied world models. InBMVC, 2025

    work page 2025

  50. [50]

    WorldOlympiad: Can Your World Model Survive a Triathlon?

    Zhao, Y ., Zhao, W., Wang, W., Zhang, Z., An, D., Liu, A., Yu, Y ., Tang, J., Wang, F., Wang, W., and Zhuang, B. Worl- dOlympiad: Can Your World Model Survive a Triathlon? arXiv preprint arXiv:2606.11129, 2026

    work page internal anchor Pith review Pith/arXiv arXiv 2026

  51. [51]

    Person moves forward; Camera turns left

    Teed, Z. and Deng, J. RAFT: Recurrent All-Pairs Field Transforms for Optical Flow.arXiv preprint arXiv:2003.12039, 2020. 17 A. Test Suite Gallery To provide a qualitative overview of the visual and action coverage in WorldOdysseyBench, we include a gallery of representative test cases in Figure 11. Each panel shows the first-frame image together with an a...

    work page arXiv 2003

  52. [52]

    camera-pose chain

    Depth-percentile Filteringretain nearest p%Far Near Cross-Segment Registration Coarse:PTY3 / FPFH+ RANSACFine:point-to-planeICPQuality-awareacceptance:fitness ≥ τᵩand Chamferimprovest … Video + GTActions⋯WW(Explore) SS(Revisit)⋯ Observation Segment(t ≤ t*) Frame-LevelPointClouds t Revisit Segment(t > t*) t Frame-LevelPointClouds t*Memory Evaluation Pipeli...

    work page

  53. [53]

    Holistic scoring accommodates smooth viewpoint and illumination changes that can confound frame-level com- parisons, while producing a single benchmark-compatible scalar without requiring an external reference-feature li- brary. G. Subject Memory Prompt For third-person memory evaluation, the Subject Memory Evaluation Prompt in Figure 27 scores video-leve...

    work page

  54. [54]

    Is there a shadow visible on the wall or floor?

    work page

  55. [55]

    If yes, does the shadow move or change in a way that is physically consistent with the camera movement and the light source position? Answer ’yes’ if the shadow appears and behaves correctly, ’no’ if the shadow is missing or behaves incorrectly. After your reasoning, conclude with exactly one line: Answer: yes or Answer: no Figure 26.Occlusion and shadow ...

    work page

  56. [56]

    Identity change: the subject becomes a different individual or category

    work page

  57. [57]

    Structural distortion: the body, anatomy, proportions, limbs, head, face, or key parts become deformed or implausible

    work page

  58. [58]

    Appearance drift: color, texture, clothing, hair, material, or style is truly rewritten

    work page

  59. [59]

    Subject disappearance: the subject becomes partly or fully invisible

    work page

  60. [60]

    subject memory score

    Quality degradation: severe blur, low resolution, diffused edges, or loss of key details makes the subject hard to identify. Important calibration rules: - Viewpoint change alone is not inconsistency. Back-to-front, front-to-back, side-to-front, or far-to-close changes are acceptable if the subject can reasonably be the same individual under the new view....

    work page