GPT-5.6 Sol Pro · judged blind run

Research physics that can still return a reward.

Original physics challenges expressed as independently gradable code checkpoints, with public, hidden, and stress tests; terse feedback; multiple attempts; and complete repair trajectories.

Research-style numerical work with executable ground truth.

We translate advanced physical models into code-level checkpoints that retain conceptual difficulty while making success reproducible, stress-testable, and useful for multi-attempt reinforcement learning.

The supplied clean-room environment contains 60 challenges, 180 independently gradable checkpoints, and 12 synthetic physics families. A challenge breaks into intermediate and main functions, each with a code template, grouped tests, reference implementation, difficulty metadata, and deterministic split.

Evaluated model · GPT-5.6 Sol Pro

GPT-5.6 Sol Pro acted across 36 held-out checkpoints. Public, hidden, and stress-test groups judged every submitted program across 71 total submissions. The result is for CritPt-Synth-Hard-v1, an independent synthetic environment.

Multi-attempt RL without leaking hidden answers.

Observation

Physics + code contract

Problem statement, function signature, family, difficulty, prior feedback, and remaining attempt budget.

Action

Executable answer

Candidate Python source defining the requested function under a constrained import and runtime policy.

Reward

Grouped verification

Syntax, policy, load, public tests, hidden tests, and stress tests produce a weighted score.

Learning loop

Delta reward

Only improvement over the episode's previous best is rewarded, preventing repeated submissions from farming reward.

GPT-5.6 Sol Pro solved ten families completely and exposed two hard clusters.

Exact solve rate83.33%30 of 36 checkpoints
Grouped cases240 / 288public + hidden + stress
Family coverage12modern physics families
Family sliceExact solvedObserved outcome
Non-Hermitian SSH, Lindblad, Floquet, Fukui12 / 12All checkpoint groups passed
Turing, MPS, birth-death, Gaussian12 / 12All checkpoint groups passed
Cosmological transfer, correlated decoder6 / 6All checkpoint groups passed
Layered neutrinos, Z2 lattice gauge0 / 6Valid code loaded; numerical results missed all test groups

The 71 submitted actions include initial observations, candidate source, concise rationale, reward, grader feedback, and terminal status. The run was offline and did not inspect expected answers, test contents, generator source, or detailed grader internals.

The unsolved clusters define the next curriculum.

Failure cluster6 checkpoints × 4 attempts
  1. Layered neutrinos: all three checkpoints loaded and complied with policy, but propagator and probability outputs remained numerically wrong after four attempts.
  2. Z2 lattice gauge: all three checkpoints produced valid functions, but partition and Wilson-loop values remained outside tolerances.
  3. Curriculum response: create smaller invariant checks, convention-disambiguation tasks, intermediate checkpoint rewards, critiques, and repaired implementations around those two mechanisms.

The environment preserves partial credit for valid, policy-compliant code even when scientific outputs fail. That makes the difference between syntax, implementation, convention, and physics errors visible to the training loop.

What a CritPt-shaped program includes.

  • Original research-level physics families with notebooks, executable JSON records, independently gradable subproblems, and difficulty metadata.
  • Gymnasium-compatible multi-attempt environment, deterministic selection, curriculum controls, and direct grading API.
  • Sandbox policy plus public, hidden, and stress tests with terse non-leaking feedback.
  • Per-attempt code, observation/reward traces, failure labels, concise rationales, critiques, and verified repairs.
  • Train, validation, and sealed test splits with manifests, seeds, replay tools, and Hugging Face-style exports.

Benchmark relationship and source boundary.

CritPt-Synth-Hard is independently authored clean-room material that follows the public task contract and broad capability shape. No official CritPt question text or solution is included, and there is no benchmark affiliation.

Primary reference: official CritPt evaluation repository. Evidence summary: JSON.

Research physics environments

Make difficult physics failures trainable.

Start a program