arXiv：cs.AI· Wei Lin·· 16 小时前
Reels-RSI：面向代码渲染解说视频的程序级自我改进框架
Reels-RSI: Program-Level Self-Improvement for Code-Rendered Explainer Videos
AI 导读
开源项目 Reels-RSI 提出在程序层面而非权重或提示词层面实现自我改进，将解说视频视为可由确定性渲染器生成帧的程序。在 111 个视频的实验中，现场学到的规则在旧帧中发现四个潜在缺陷、新帧零缺陷，并作为弱导演的守卫（未守护 30 次违规对 5 次）。研究还揭示首轮改写 11-1 的判定源于投票设计中的位置偏差（评委 62-81% 选第二个视频）。
正文
Abstract:Explainer videos -- a worked geometry problem, why the sky is blue, how a dynasty fell -- are usually one more target for pixel video generation. We argue they are a different kind of artifact: a program (script, storyboard and per-frame animation code) that a deterministic renderer turns into frames. A program can be read, diffed, tested and audited; pixels can only be watched. Reels-RSI is an open-source reel director for any agent harness; its self-improvement happens at the program level rather than in weights or prompts. Every revision pass is logged with a code snapshot; a retrospective pairs each check finding with the diff that fixed it and proposes lessons; a lesson that a static check can catch is compiled into a rule with a must-fail and a must-pass example, so the memory is unit-tested; and a rule learned today runs over everything made before it. A bounded, benchmark-gated step lets an agent rewrite the director's playbook, kept only when blind pairwise judgments on held-out topics prefer it; nothing enters memory without a human decision. Across 111 reels -- a field study, two playbook evolutions, a 2x2 memory ablation, a weaker director, a 48-reel learning curve and a judge-reliability study -- we find that the rule learned in the field caught four latent defects in frames authored before it existed and none after; that rules are an audit for a strong director (eight unguarded reels, no violation) and a guard for a weaker one (thirty violations unguarded versus five); that lessons shorten revision loops but do not compound or move the judge; and that the first rewrite's 11--1 verdict was a position bias in our vote design (judges pick the second-shown reel 62--81% of the time), masking scripts that traded thirteen facts for craft. We document four failure modes of the preference signal, which check caught each, and the fixes. Code, rules, lessons and data are public.
| Comments: | 19 pages, 4 figures, 9 tables. Code, rules, lessons, data and evolve reports: this https URL |
| Subjects: | Artificial Intelligence (cs.AI); Graphics (cs.GR); Multimedia (cs.MM); Software Engineering (cs.SE) |
| Cite as: | arXiv:2610.14495 [cs.AI] |
| (or arXiv:2610.14495v1 [cs.AI] for this version) | |
| https://doi.org/10.48550/arXiv.2610.14495 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Wei Lin [view email]
[v1]
Sat, 10 Oct 2026 02:46:44 UTC (404 KB)
来源：arXiv：cs.AI · arxiv.org