H1r0t0 Ayanokoji

hiroto101

AI & ML interests

None yet

Recent Activity

updated a collection 3 days ago

NSFW

liked a model 3 days ago

NSFW-API/NSFW_Wan_1.3b

reacted to Chris4K's post with 🔥 5 days ago

Can a network whose trunk is only XOR, table lookups and integer adds learn Connect Four? Partly. I trained a GF(2⁸) integer model (8.4M params, CPU-only) on 320k games labelled by a negamax search teacher. What it learned: • agrees with the teacher on 59.6% of training / 41.0% of board-disjoint holdout positions • beats a random player 85.2% of the time What it didn't: • loses to a 1-ply "win now / block now" rule (wins 37.2%) • loses to 2-ply lookahead (wins 24.3%) Agreeing with a teacher is not the same as playing well, and this release measures both. Caveat up front: the float32 read-out is 99.6% of the parameters. The trunk is float-free; the full model isn't yet. The repo also documents six bugs that each produced a plausible but wrong number. In one, 41% of the supervision was silently all-zero. 🎮 Play: https://huggingface.co/spaces/Chris4K/connect-four-gf256 (Space) 🧠 Model: https://huggingface.co/Chris4K/connect-four-gf256 📦 Dataset: https://huggingface.co/Chris4K/connect-four-gf256 320k games + per-ply targets, MIT 💻 Code + paper: https://github.com/Ckal/connect-four-gf256 Feedback welcome, especially: which baseline would convince you?

View all activity

Organizations

None yet

updated a collection 3 days ago

NSFW

Collection

2 items • Updated 3 days ago

liked a model 3 days ago

NSFW-API/NSFW_Wan_1.3b

Updated Jun 25, 2025 • 1.58k

reacted to Chris4K's post with 🔥 5 days ago

Post

3,018

Can a network whose trunk is only XOR, table lookups and integer adds learn Connect Four?

Partly. I trained a GF(2⁸) integer model (8.4M params, CPU-only) on 320k games labelled by a negamax search teacher.

What it learned:
• agrees with the teacher on 59.6% of training / 41.0% of board-disjoint holdout positions
• beats a random player 85.2% of the time

What it didn't:
• loses to a 1-ply "win now / block now" rule (wins 37.2%)
• loses to 2-ply lookahead (wins 24.3%)

Agreeing with a teacher is not the same as playing well, and this release measures both.

Caveat up front: the float32 read-out is 99.6% of the parameters. The trunk is float-free; the full model isn't yet.

The repo also documents six bugs that each produced a plausible but wrong number. In one, 41% of the supervision was silently all-zero.

🎮 Play: Chris4K/connect-four-gf256 (Space)
🧠 Model: Chris4K/connect-four-gf256
📦 Dataset: Chris4K/connect-four-gf256 320k games + per-ply targets, MIT
💻 Code + paper: https://github.com/Ckal/connect-four-gf256

Feedback welcome, especially: which baseline would convince you?