byte-vortex's picture
Update logbook: Repro - NonZero: Interaction-Guided Exploration for Multi-Agent Monte Carlo Tree Search
7e6abbe verified
|
Raw
History Blame Contribute Delete
8.04 kB

Executive summary


Partial reproduction of NonZero. Of the paper's 6 claims, Claim 1's core mathematical formulation was verified exactly (finite-difference identity holds to floating-point precision), and the surrogate model was shown to learn true reward structure at its theoretical noise ceiling (R2=0.90) on a toy-scale test. Claim 4's supporting infrastructure (environment, baseline training pipeline, and a compiling but not-yet-wired C++ patch for NonUCT) was built and validated at the component level, but no end-to-end NonZero-vs-baseline comparison was completed, so Table 1's reported numbers were neither reproduced nor falsified. Claims 2-3 (theoretical proofs) and Claims 5-6 (SMAC/SMACv2) were not attempted: the former due to time constraints, the latter because the paper's own training recipe requires A100-class GPU hardware unavailable on free-tier compute. All work was done on a single Kaggle T4 GPU, free tier, over approximately 3 hours.

Scope & cost

This reproduction Full replication
Scope Claim 1 math + Claim 4 infrastructure (env, baseline, surrogate) All 6 claims incl. SMAC/SMACv2
Hardware 1x Kaggle T4 (free tier) Cluster of NVIDIA A100/A6000 GPUs
Compute time ~3 hours Many GPU-days
Cost $0 Thousands of dollars
Outcome Claim 1 verified (toy scale); Claim 4 infrastructure validated, not completed; Claims 2-3, 5-6 not attempted

<!DOCTYPE html>
<html><head><meta charset="utf-8">
<style>
  body { font-family: -apple-system, Helvetica, Arial, sans-serif; margin: 0;
         padding: 32px; background: #0f1115; color: #f2f2f2; }
  h1 { font-size: 26px; margin-bottom: 4px; }
  h2 { font-size: 15px; font-weight: 400; color: #9aa; margin-top: 0; }
  .grid { display: grid; grid-template-columns: 1fr 1fr 1fr; gap: 16px; margin-top: 20px; }
  .card { background: #1a1d24; border: 1px solid #2a2e38; border-radius: 8px; padding: 14px; }
  .card h3 { margin: 0 0 6px; font-size: 13px; color: #7fd1ff; }
  .card p { margin: 0; font-size: 12px; line-height: 1.4; color: #d8d8d8; }
  .status { display: inline-block; padding: 2px 8px; border-radius: 10px; font-size: 10px;
            font-weight: 600; margin-bottom: 6px; }
  .s-verified { background: #1e4d2b; color: #7ee2a0; }
  .s-partial  { background: #4d3f1e; color: #f0c674; }
  .s-reviewed { background: #1e3a4d; color: #7ec8f0; }
  .s-none     { background: #3a1e1e; color: #f08080; }
  footer { margin-top: 20px; font-size: 11px; color: #888; }
  a { color: #7fd1ff; }
</style></head>
<body>
  <h1>NonZero: Interaction-Guided Exploration for Multi-Agent MCTS</h1>
  <h2>Reproduction notes -- ICML 2026 Agent-Repro Challenge &middot; <a href="https://openreview.net/forum?id=Jh6gq9QsFa">OpenReview Jh6gq9QsFa</a> &middot; <a href="https://arxiv.org/abs/2605.00751">arXiv:2605.00751</a></h2>
  <div class="grid">
    <div class="card"><span class="status s-verified">VERIFIED (toy)</span>
      <h3>Claim 1 -- asinh-GLM surrogate</h3>
      <p>Finite-difference identity holds to floating-point precision (~1.19e-7). R2=0.90 on held-out reward fitting, matching the environment's noise ceiling.</p></div>
    <div class="card"><span class="status s-reviewed">REVIEWED</span>
      <h3>Claim 2 -- Regret bound (Thm 3.5)</h3>
      <p>Proof (App. A.2-A.5) is internally consistent, standard techniques. Analyzes an idealized continuous-update proxy, not the literal discrete Algorithm 1 mechanism.</p></div>
    <div class="card"><span class="status s-partial">TOY ILLUSTRATION</span>
      <h3>Claim 3 -- Efficiency separation (Thm 3.7)</h3>
      <p>Separation ratio measured n=3..8: 160x -> 7.19M x, growing 44,939x while the full action space only grew 32,768x -- consistent with exponential-in-n advantage.</p></div>
    <div class="card"><span class="status s-partial">PARTIAL</span>
      <h3>Claim 4 -- MatGame 8-agent/10-action</h3>
      <p>Env + baseline pipeline + compiling C++ patch validated; toy HF Job shows the mechanism beating random search by ~43%. Table 1's exact numbers not reproduced.</p></div>
    <div class="card"><span class="status s-none">NOT ATTEMPTED</span>
      <h3>Claims 5 &amp; 6 -- SMAC / SMACv2</h3>
      <p>Paper's training recipe requires A100-class GPU; free-tier Kaggle T4/P100 cannot produce comparable numbers.</p></div>
    <div class="card"><span class="status s-verified">$0 / ~4 hrs</span>
      <h3>Scope &amp; cost</h3>
      <p>1x Kaggle T4 (free) + HF Jobs (cpu-basic), ~3-4 hours total, vs. a full replication needing an A100/A6000 cluster over many GPU-days.</p></div>
  </div>
  <footer>Full logbook: <a href="https://huggingface.co/spaces/byte-vortex/nonzero-repro">huggingface.co/spaces/byte-vortex/nonzero-repro</a></footer>
</body></html>

Partial reproduction of NonZero. Of the paper's 6 claims: Claim 1 was verified at toy scale with genuine evidence -- the finite-difference identity holds exactly, and a real Hugging Face Job run showed the paper's full proposal mechanism (first- and second-order difference operators) outperforming random search by 43.4% under equal evaluation budget. Claim 2 (regret bound, Theorem 3.5) was reviewed against the actual proof in the paper (Appendix A.2-A.5): the proof strategy and algebra are internally consistent and use standard techniques, but it analyzes an idealized continuous-update proxy for NonUCT rather than the literal discrete finite-difference mechanism in Algorithm 1, and its complexity constants aren't derived for the paper's actual hypernetwork architecture -- so the proof holds as written but leaves an unverified gap between theory and the described algorithm. Claim 3 (efficiency separation, Theorem 3.7) has a toy numerical illustration extended to n=3..8 agents: the guided-search-budget-vs-full-space separation ratio grew from 160x (n=3) to ~7.19 million x (n=8), a ~44,939x growth while the full joint-action space itself only grew 32,768x over the same range -- consistent with the theorem's claimed exponential-in-n advantage, though this is a qualitative illustration, not a check of the theorem's exact constants. Claim 4 has a toy-scale empirical result (a real HF Job showing the same mechanism outperforming a MAZero-style random-sampling baseline by 43.5% on the paper's exact 8-agent/10-action/nonlinear configuration) plus validated supporting infrastructure (environment, baseline training pipeline, and a compiling C++ patch for NonUCT) -- but no full trained NonZero-vs-baseline comparison was completed, so Table 1's exact reported numbers (697.1 vs 672.3) were neither reproduced nor falsified. Claims 5-6 (SMAC/SMACv2) were not attempted: the paper's own training recipe requires A100-class GPU hardware unavailable on free-tier compute.

Scope & cost

This reproduction Full replication
Scope Claim 1 (verified, toy) + Claim 2 (proof reviewed) + Claim 3 (extended toy illustration) + Claim 4 (toy empirical result + infrastructure) All 6 claims incl. SMAC/SMACv2
Hardware Kaggle T4 (free tier) + Hugging Face Jobs (cpu-basic) Cluster of NVIDIA A100/A6000 GPUs
Compute time ~3-4 hours Many GPU-days
Cost $0 Thousands of dollars
Outcome Claim 1 verified (toy scale, real evidence); Claim 2 proof reviewed (gap noted, not a numerical claim); Claim 3 toy illustration extended (n=3..8); Claim 4 toy result + infrastructure validated, full comparison not completed; Claims 5-6 not attempted (compute-bound)