[codex] Add July 2026 world model curation updates

[iChubai]

Summary

• Refresh the README recency marker to July 5, 2026.
• Add 19 strictly screened world-model papers/resources from late June and early July 2026 across the existing taxonomy.
• Keep entries distributed by OpenEnvision category: driving occupancy, robot manipulation, navigation, WAM/VLA, video world-model backbones, JEPA, MBRL/planning/evaluation, surveys, and benchmarks.

Curation Notes

Included papers have a direct world-model mechanism such as action-conditioned prediction, latent/3D state rollout, closed-loop simulation, planning/evaluation through imagined futures, or benchmark/survey coverage for those systems. I prioritized entries with official project/code, benchmark signal, venue signal, or clear inclusion in recent external curated lists.

Filtered out adjacent or low-signal candidates that were mostly generic video generation, retail/deployment tooling, broad theory without a clear actionable world-model contribution, or duplicates already covered by the README.

Primary sources checked include arXiv recent submissions plus updates from knightnemo/Awesome-World-Models, LMD0311/Awesome-World-Model, leofan90/Awesome-World-Models, Li-Zn-H/AwesomeWorldModels, gracezhao1997/Awesome-Video-World-Models-with-AR-Diffusion, and ziqihuangg/Awesome-From-Video-Generation-to-World-Model.

Checks

• node scripts/check-arxiv-duplicates.mjs README.md
• git diff --check

[iChubai]

Closing this duplicate draft; the July 5 increment was folded into the existing cumulative curation PR #8.