Matthew Siper

Computer scientist obsessed with markets, Founder builder @the_nof1

[Matthew Siper @MatthewSiper]

1/6 🧵

How do you train a language model to become a self-adaptive, long-horizon optimization engine for executable trading policy search?

In our new paper, "The Time Value of Evolution," we introduce Lineage-Value Policy Gradients, or LVPG, a new long-horizon RL technique for adaptive evolutionary search.

[Matthew Siper @MatthewSiper]

@togelius Amen

[Matthew Siper @MatthewSiper]

@utheprodigyn @schulzb589 @Amidos2006 @togelius Yes give it a read let me know what you think 😂

[Matthew Siper @MatthewSiper]

New work on programmable cellular automata by @Amidos2006 @utheprodigyn @togelius and myself!

[Ahmed Khalifa @Amidos2006]

Cellular automata are powerful but painful to author. Rules are a lookup table; neural CAs are more expressive but a black box.

What if each CA rule was just code that you can read?

Introducing Programmable Cellular Automata (PCA)

[Julian Togelius @togelius]

.@Amidos2006 presenting our work (led by @MatthewSiper ) on Procedural Content Metageneration via Continual Abstraction Discovery here at @ieee_cog

[Ravid Shwartz Ziv @ziv_ravid]

This week on The Information Bottleneck we're talking with Julian Togelius (@togelius ) 🥳🥳🥳

Julian directs the Game Innovation Lab at NYU and co-founded https://t.co/SRVCYsLB2x. For years he's worked on AI and games from both directions: using AI to generate game content and test games, and using games to actually measure what AI can do. Lately he's been digging into why LLMs that write code so well still fail badly at playing games, and what that gap tells us about spatial reasoning and learning in current models.

We'll get into game playing as a benchmark, procedural content generation, open-endedness, and where he thinks the AGI conversation goes wrong.

What should we ask him? Drop your questions in the comments

[Matthew Siper @MatthewSiper]

Wild

[will brown @willcb]

trained on prime intellect lab :)

this is nuts my mind is kinda blown

[Matthew Siper @MatthewSiper]

5/5

CAD turns program search into a process whose vocabulary changes during the run. Later programs inherit callable abstractions discovered from earlier high-fitness programs.

Accepted at IEEE CoG 2026.

With @Amidos2006 @togelius.

Website: https://t.co/jHHobV7t5Q

Paper: https://t.co/ZViYgA9jsP

[Matthew Siper @MatthewSiper]

🧵 1/5

What happens when a language model doesn't just write code, but dynamically creates its own expert API mid-run?

Instead of evolving individual game levels, our new paper evolves entire generator programs using an LLM as the mutation and crossover operator.

[Matthew Siper @MatthewSiper]

4/5

CAD raised mean final-best in all comparisons across Sokoban, Zelda, Lode Runner, and Dangerous Dave.

The endpoint favored CAD whether search began with no helper library or a fixed hand-written expert API.

[Arda Uzunoğlu @aardauzunoglu]

🚨 new preprint: Information Abundance Paradox

TL;DR: Longer context can weaken parametric learning in LLMs.

We propose the Information Abundance Paradox: when relevant information is abundant in the training context, the model has less incentive to internalize that information in its parameters.

[Matthew Siper @MatthewSiper]

Hi @DefiApes we use walk-forward for this. 10 years of data per asset is split into folds. Within a fold is a training period (hill climbing), a validation period (the model can only parameter tune here no logic updates), and test period (completely sealed and only evaluated after search is completed for that fold). 10-day embargos between periods within a fold help prevent leakage. There are several limitations here in terms of the ability to find purely general non-overfit policies (eg the since objective function for the reported test values is validation sharpe then the test sharpe is partly predicated on how autocorrelated the data is / how similar the test period is to the validation period). However, we have research coming that addresses policy robustness in a novel way that I’m excited to share soon.

[Matthew Siper @MatthewSiper]

1/6 🧵

How do you train a language model to become a self-adaptive, long-horizon optimization engine for executable trading policy search?

In our new paper, "The Time Value of Evolution," we introduce Lineage-Value Policy Gradients, or LVPG, a new long-horizon RL technique for adaptive evolutionary search.

[Jay A @jay_azhang]

At Nof1 we believe the next breakthrough in AI after reasoning is adaptation

If reasoning is the ability to detect patterns and leverage them, adaptation is anticipating how those patterns will change

This is a crucial capability for real world AI, and current LLMs struggle with it. We've run thousands of live market experiments and no amount of context, prompting, or harness engineering has worked. So, we've been pushing our research deeper down the stack

In this paper, we trained models that can better adapt in dynamic environments like markets. They learn to value an action by the futures it opens up, not just its immediate payoff

It's an early step, but we're building towards forever learners that are default forward-looking, rather than static and backward looking

[Matthew Siper @MatthewSiper]

6/6

By reframing evolutionary search as long-horizon RL, we train a language model to self-correct and autonomously evolve robust policies.

This work was done with my co-authors @togelius @Amidos2006.

Preprint: https://t.co/PAKJBnOrBR

#ReinforcementLearning #EvolutionaryComputation #LLMs #MachineLearning #AlgorithmicTrading

[Matthew Siper @MatthewSiper]

5/6

Stronger finite-budget search:

Across 90 matched paired runs evolving quantitative trading policies, LVPG increased mean sealed-test Sharpe from 0.86 to 1.32, while peak sealed-test Sharpe reached 4.13.