LLM 零样本权重空间编程:跨越难度阶梯的双人游戏网络构建研究
Zero-Shot Weight-Space Programming Across a Difficulty-Graded Ladder of Two-Player Games
AI 导读
一篇 arXiv 论文研究大语言模型能否零样本完成权重空间编程,即不靠梯度下降和训练数据、按显式规范手工设定前馈网络全部权重。作者设计四级难度阶梯(Nim、井字棋、两种尺寸 Connect Four、国际象棋残局),构建并独立验证 10 个网络,其中 6 个获得完整的非统计正确性证明。
正文 · 原文
Abstract:Weight-space programming sets every weight of a feed-forward network by hand, from an explicit specification of the desired computation, with no gradient descent and no training data anywhere in the construction. We ask how far a large language model can carry this practice, zero-shot, across games of deliberately increasing difficulty. We design a four-rung ladder (Nim, tic-tac-toe, two sizes of Connect Four, and a chess endgame) spanning four axes of difficulty, and build a second, independent network at each rung, using a structurally unrelated game at matched difficulty, to separate what a rung's finding says about the difficulty level from what it says about the one game used to measure it. Ten networks are built and independently verified. Six admit a complete, non-statistical proof of correctness against an independent oracle. The chess-endgame rung is exact where correctness is cheap, move legality and, in one network, an exact one-ply mate detector, and collapses where it is expensive, composing a multi-move mating plan, from 84.1-100% success at the one-ply tactics it was built to get right down to 0.57-9.13% in full adversarial play; its replicate shows this collapse is an order of magnitude less severe under a shorter required plan and exact rather than heuristic safety checks. The standard-sized rung confirms, by holding hand-set evaluation weights fixed and adding only shallow search, that the remaining gap is an architectural ceiling rather than a matter of engineering effort, and this exact collapse reproduces across two structurally unrelated games. We release all ten networks, their independent oracles, construction-cost measurements, and a cold-prompt protocol for a future cross-model comparison
| Subjects: | Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2611.09271 [cs.AI] |
| (or arXiv:2611.09271v3 [cs.AI] for this version) | |
| https://doi.org/10.48550/arXiv.2611.09271 arXiv-issued DOI via DataCite |
Submission history
From: Eduardo C. Garrido-Merchán [view email]
[v1]
Fri, 6 Nov 2026 10:26:48 UTC (34 KB)
[v2]
Mon, 9 Nov 2026 15:52:19 UTC (34 KB)
[v3]
Tue, 17 Nov 2026 09:46:57 UTC (42 KB)
来源:arXiv cs.AI · arxiv.org