DeepSeek V4.1 Flash 1017 can also do Weight-Space Programming
retained text
Inspired by [Weight-Space Programming](https://rikiyat.github.io/blog/weight-space-programming/), I tested whether DeepSeek V4.1 Flash 1017 could figure out what a small neural network does by reading its weights. I fed it four anonymized RL checkpoints: **51k parameters each, identical architectures**, with all weights rendered as text. No metadata, network source code or execution. It received the architecture outline and four candidate tasks. **Results:** CartPole ✅ Flappy ✅ Catch ✅ BallBeam ❌ It picked out patterns consistent with relative bird/gap position, paddle/ball position and angle-based balancing. It confused BallBeam with CartPole an interesting mistake because both share the same observation layout and similar balancing controls. **Very preliminary: one run per checkpoint.** This suggests some ability to interpret weights, but four trials don’t establish reliability. These were API runs with maximum reasoning effort, consuming \~1.58M tokens total. I may try more replicates and independently trained checkpoints, plus include the network code with task-revealing details removed. The poor LLM had to guess quite a lot it even assumed ReLU where the network uses `tanh` 😅 **Code, checkpoints and run logs:** [https://github.com/galatolofederico/wsp](https://github.com/galatolofederico/wsp)
document fetched 2026-11-04 00:41 · canonical https://reddit.com/r/LocalLLaMA/comments/1xn2i0n/deepseek_v41_flash_1017_can_also_do_weightspace