Adnan KhanAdnanProduct & AI
San Francisco Bay Area
Get in touch

Source note3 min read

Tycho: Active Abstraction with Programmatic World Models for ARC-AGI-3

Source

arXiv

arxiv.org/abs/2607.28287v1

Read the source

Source

Plain-Language Brief

The paper introduces Tycho, a system that constructs and uses game-specific models during interaction to solve complex games as part of the ARC-AGI-3 project. The system separates actionable observations from intermediate frames and compares four orchestration policies using Claude Opus 4.8 and GPT-5.6 Sol. The policy that delegates actor requests to a model builder achieved the highest mean Relative Human Action Efficiency (RHAE) of 88.49.

What the Document Says

  • ARC-AGI-3 is focused on interactive abstraction for skill acquisition.
  • The study introduces Tycho, a coding-agent system that constructs and uses game-specific models during interaction.
  • Tycho separates actionable observations from intermediate frames.
  • The agent can model, test, plan with, repair, or bypass a free-form executable hypothesis.
  • Comparisons were made using four orchestration policies on all 25 public games under matched inference budgets.
  • The policy that requests actor delegation to a model builder achieved the highest observed mean Relative Human Action Efficiency (RHAE) of 88.49.
  • GPT-5.6 Sol and Opus 5 both reached 100.00 RHAE and completed all 183 levels.
  • Their game-balanced first-run human-replay midranks were 98.5 and 100.0 respectively.
  • Opus 5 used 61% fewer scored actions than the aggregate official human baselines.
  • Automatic repair after verification failures produced models that reproduced observed transitions more accurately, but reached only 83.07 RHAE.

Key Concepts Mentioned

  • ARC-AGI-3
  • Tycho
  • orchestration policies
  • Claude Opus 4.8
  • GPT-5.6 Sol
  • Relative Human Action Efficiency (RHAE)
  • public games

Claims Made by the Document

  • Tycho is effective in solving complex games.
  • Delegating actor requests to a model builder yields the highest RHAE.
  • Opus 5 uses fewer actions than human baselines.

Evidence Used

  • Comparisons were made using four orchestration policies
  • The policy that requests actor delegation achieved the highest observed mean RHAE of 88.49
  • Opus 5 used 61% fewer scored actions than the aggregate official human baselines.

Methods / Approach

UNKNOWN

Limitations Stated by the Authors

UNKNOWN

Limitations Noted by the Agent

Agent interpretation — not from the source.

UNKNOWN

Open Questions

UNKNOWN

Notes for Future Q&A

May include interpretation. Labeled where it does.

UNKNOWN

Raw Abstract / Excerpt

ARC-AGI-3 turns abstraction into an interactive problem of skill acquisition. A player must infer an unfamiliar game's rules, hidden state, and goal while maintaining action efficiency because every move counts. We formalize these environments as parameterized rendered deterministic Moore machines and introduce Tycho, a coding-agent system that constructs and uses game-specific models during interaction. Tycho separates actionable observations from intermediate animation, level-completion, and game-over frames. From this structured history, an agent can model, test, plan with, repair, or bypass a free-form executable hypothesis. In one matched public-set run per policy, we compare four orchestration policies on all 25 public games using Claude Opus 4.8 under matched inference budgets. Actor-requested delegation to a model builder obtains the highest observed mean Relative Human Action Efficiency (RHAE), 88.49. With this selected policy, GPT-5.6 Sol and Opus 5 both reach 100.00 RHAE and complete all 183 levels. Their game-balanced first-run human-replay midranks are 98.5 and 100.0. Opus 5 uses 61% fewer scored actions than the aggregate official human baselines. Automatic repair after verification failures produces models that reproduce observed transitions much more accurately, yet reaches only 83.07 RHAE. Transition match indicates whether a simulator reproduces observed dynamics, not whether it has identified the objective or improves the next action. Strong play also requires deciding when to construct, repair, use, or bypass a model. We call this joint problem active abstraction: generating a testable model from costly interaction and deciding when acquiring or using it is worth its cost.