Repository navigation
Paper review 2026-08-27: 2 candidates #367
Description
Reply with commands, one per line:
/approve all · /approve 1,3-5 · /reject 2 · /edit 3 category=world_terminal tags=benchmark (edit implies approve; tags=- clears tags). Valid category keys: see taxonomy.json.
1. Code World Model: Coding Agent as World Brain
Yiwen Chen, Guosheng Lin, Chi Zhang · arXiv 2026/08 · paper
proposed: world_game · tags: none
Introduces Code World Model, where a coding agent generates and executes code to maintain persistent, rule-consistent world state, which is compiled into a proxy video conditioning a video model for high-fidelity rendering of interactive gameplay and real-world scenes.
Code serves as the executable policy for simulating and evolving an interactive world rather than as a delivered software artifact, placing it in the agency/world_game branch.
2. From Blind Edits to Verified Repair: Building Trustworthy User-Side LLM Agents for Web Accessibility
Lily Bundgaard Wanscher, Markus Heidemann Lorensen, Mohammed Ammad Shafiq, et al. · arXiv 2026/07 · paper
proposed: world_browser · tags: benchmark
A browser extension agent that generates and injects additive CSS to repair web accessibility violations on live pages, paired with a verified audit-inject-verify loop and benchmark to prevent regressions.
Agent writes and executes CSS code to act on live web pages, fitting world_browser's code-driven web action criterion.
machine payload (do not edit)
[{"paper": {"id": "2608.25927", "title": "Code World Model: Coding Agent as World Brain", "authors": ["Yiwen Chen", "Guosheng Lin", "Chi Zhang"], "venue": "arXiv 2026/08", "category": "", "published": "2026-08-26", "links": {"paper": "https://arxiv.org/abs/2608.25927", "github": "", "website": ""}}, "category": "world_game", "tags": [], "summary": "Introduces Code World Model, where a coding agent generates and executes code to maintain persistent, rule-consistent world state, which is compiled into a proxy video conditioning a video model for high-fidelity rendering of interactive gameplay and real-world scenes.", "reason": "Code serves as the executable policy for simulating and evolving an interactive world rather than as a delivered software artifact, placing it in the agency/world_game branch.", "source": "crawl"}, {"paper": {"id": "2608.24913", "title": "From Blind Edits to Verified Repair: Building Trustworthy User-Side LLM Agents for Web Accessibility", "authors": ["Lily Bundgaard Wanscher", "Markus Heidemann Lorensen", "Mohammed Ammad Shafiq", "Mahyar Tourchi Moghaddam", "Mina Alipour"], "venue": "arXiv 2026/07", "category": "", "published": "2026-07-26", "links": {"paper": "https://arxiv.org/abs/2608.24913", "github": "", "website": ""}}, "category": "world_browser", "tags": ["benchmark"], "summary": "A browser extension agent that generates and injects additive CSS to repair web accessibility violations on live pages, paired with a verified audit-inject-verify loop and benchmark to prevent regressions.", "reason": "Agent writes and executes CSS code to act on live web pages, fitting world_browser's code-driven web action criterion.", "source": "crawl"}]
Activity