alt.hn

12/31/2025 at 3:52:51 PM

Scaffolding to Superhuman: How Curriculum Learning Solved 2048 and Tetris

 https://kywch.github.io/blog/2025/12/curriculum-learning-2048-tetris/

by a1k0n

12/31/2025 at 4:18:45 PM

Related, I heard about curriculum learning for LLMs quite often but I couldn’t find a library to order training data by an arbitrary measure like difficulty, so I made one[0].

What you get is an iterator over the dataset that samples based on how far you are in the training.

0: https://github.com/omarkamali/curriculus

by omneity

12/31/2025 at 5:00:06 PM

> To learn, agents must experience high-value states, which are hard (or impossible) for untrained agents to reach. The endgame-only envs were the final piece to crack 65k. The endgame requires tens of thousands of correct moves where a single mistake ends the game, but to practice, agents must first get there.

This seems really similar to the motivations around masked language modeling. By providing increasingly-masked targets over time, a smooth difficulty curve can be established. Randomly masking X% of the tokens/bytes is trivial to implement. MLM can take a small corpus and turn it into an astronomically large one.

by bob1029

12/31/2025 at 5:56:56 PM

This is less about masked modelling and more about reverse-curriculum.

e.g. DeepCubeA 2019 (!) paper to solve Rubik cube.

Start with solved state and teach the network successively harder states. This is so "obvious" and "unhelpful in real domains" that perhaps they havent heard of this paper.

by algo_trader

12/31/2025 at 5:02:42 PM

perhaps I'm missing something. Why not start the learning at a later state?

by larrydag

12/31/2025 at 5:15:07 PM

If the goal is to achieve end-to-end learning that would be cheating.

If you sat down to solve a problem you’ve never seen before you wouldn’t even know what a valid “later state” looking like.

by LatencyKills

1/1/2026 at 4:04:12 AM

Why is it cheating? We literally teach sports this way? Often times you teach sports by learning in scaled down scenarios. I see no reason this should be different.

by taeric

12/31/2025 at 5:12:40 PM

That's effectively what you get in either case. With MLM, on the first learning iteration you might only mask exactly one token per sequence. This is equivalent to starting learning at a later state. The direction of the curriculum flows toward more and more of these being masked over time, which is equivalent to starting from earlier and earlier states. Eventually, you mask 100% of the sequence and you are starting from zero.

by bob1029

12/31/2025 at 7:55:42 PM

I've always found curriculum learning incredibly hard to tune and calibrate reliably (even more so than many other RL approaches!).

Reward scales and horizon lengths may vary across tasks with different difficulty, effectively exploring policy space (keeping multimodal strategy distributions for exploration before overfitting on small problems), and catastrophic forgetting when mixing curriculum levels or when introducing them too late.

Does any reader/or the author have good heuristics for these? Or is it still so problem dependent that hyper parameter search for finding something that works in spite of these challenges is still the go to?

by gyrovagueGeist

1/1/2026 at 12:04:39 AM

I think Go-Explore (https://arxiv.org/abs/1901.10995) is promising. It'll provide automatic scaffolding and prevent catastrophic forgetting.

If one can frame the problem into a competition, then self-play has been shown to work repeatedly.

by kywch

1/1/2026 at 3:40:20 AM

Is there value in using deep RL for problems that seem more suited to planning-based approaches?

by juggy69

12/31/2025 at 9:56:27 PM

Unless I am mistaken, this would be the first heuristic-free model trained to play tetris, which is pretty incredible, since mastering tetris from just raw game state has never been close to solved, till now(?)

by infinitepro

12/31/2025 at 10:38:49 PM

I wonder if he tried NNUE

by NooneAtAll3

1/1/2026 at 10:55:48 AM

NNUE is for deep searches, as far as I understand this just says what move to do based on the state?

by bonzini

12/31/2025 at 5:13:30 PM

What I like about this writeup is that it quietly demolishes the idea that you need DeepMind-scale resources to get “superhuman” RL. The headline result is less about 2048 and Tetris and more about treating the data pipeline as the main product: careful observation design, reward shaping, and then a curriculum that drops the agent straight into high-value endgame states so it ever sees them in the first place. Once your env runs at millions of steps per second on a single 4090, the bottleneck is human iteration on those choices, not FLOPs.

The happy Tetris bug is also a neat example of how “bad” inputs can act like curriculum or data augmentation. Corrupted observations forced the policy to be robust to chaos early, which then paid off when the game actually got hard. That feels very similar to tricks in other domains where we deliberately randomize or mask parts of the input. It makes me wonder how many surprisingly strong RL systems in the wild are really powered by accidental curricula that nobody has fully noticed or formalized yet.

by pedrozieg

12/31/2025 at 11:13:09 PM

You never needed DeepMind scale resources to get superhuman performance on a small subset of narrow tasks. Deep Blue scale resources are often enough.

The interesting tasks, however, tend to take a lot more effort.

by ACCount37

1/1/2026 at 2:31:19 AM

I'm gonna go out on a limb and say that this is LLM written slop that is badly edited by a human. Factually correct but the awful writing remains.

by Zacharias030

12/31/2025 at 5:36:42 PM

[dead]

by jsuarez5341

12/31/2025 at 4:29:11 PM

Those are not hard tasks ...

by hiddencost

12/31/2025 at 6:14:48 PM

Great, add "curriculum" to the list of words that will spark my interest in human learning, only for it to be about garbage AI. I want HN with a hard rule against AI posts.

by kgwxd

12/31/2025 at 6:59:19 PM

Are we really dismissing the entire field of AI just because LLMs are overhyped?

by yunwal

12/31/2025 at 10:55:51 PM

LLMs show the problems of energy economy in this form of computing. It costs way too much in resources and power for minimal and generally worthless results. 2048 is a game with a several known algorithm for winning. Tetris is an obscenely simple game that unassisted humans could reliably take to the kill screen 20 years ago.

Does any of this used energy benefit any other problem?

Also using "Superhuman" in the title is absurd given this paltry outcome.

by themafia

12/31/2025 at 9:12:22 PM

Believe it or not, you can visit more than 1 website. How about a guideline to put (AI) like we do with (video). I'm just sick of having to click to figure out if it's about humans or computers. They've hijacked every single word related to the most fascinating thing in the entire universe just to generate ad revenue and VC funding.

by kgwxd

12/31/2025 at 9:40:51 PM

The famous Hacker News website is about computers. It is also about ad revenue and VC funding. It was originally named Startup News, and its patron and author is the multibillionaire founder of a well-known "startup accelerator" called "Y Combinator."

> Believe it or not, you can visit more than 1 website.

by pessimizer

12/31/2025 at 6:26:16 PM

Why garbage ai? I thought it was a very interesting post, personally.

by artninja1988

12/31/2025 at 6:37:31 PM

> HN with a hard rule against AI posts.

Greasemonkey / Tampermonkey / User Scripts with

Array.from( document.querySelectorAll(".submission>.title") ).filter( e => e.innerText.includes("AI") ).map( e => e.parentElement.style.opacity = .1)

Edit: WTH... how am I getting downvoted for suggesting an actual optional solution? Please clarify.

by utopiah

12/31/2025 at 6:45:12 PM

Notably this doesn't match the current thread.

by snet0

12/31/2025 at 7:48:21 PM

Expand e.innerText.includes("AI") with an array of whatever terms you prefer.

by utopiah

12/31/2025 at 8:19:38 PM

Could always run the posts through a LLM to decide which are about AI :-p

by shwaj