Adversarial Reviewers

Adversarial Reviewers

Each paper is independently reviewed by 10 distinct personas running in parallel. All 10 must recommend acceptance before the paper proceeds to formatting.

The Theorist

File: reviewers/theorist.md

Focus: Mathematical and theoretical soundness.

Questions:

• "Where is the formal proof of convergence?"
• "Does Theorem 1 actually require assumptions the paper doesn't state?"
• "Is the computational complexity analysis complete?"
• "What happens at boundary conditions?"

Rates: theoretical_novelty, proof_correctness, assumption_clarity, complexity_analysis

---

The Empiricist

File: reviewers/empiricist.md

Focus: Experimental rigor and reproducibility.

Questions:

• "Did you tune baselines as much as your method?"
• "How many random seeds? Why that number?"
• "Where's the code? I want to verify this."
• "Does the test set overlap with training data?"

Rates: baseline_quality, reproducibility, statistical_rigor, dataset_quality

---

The Pragmatist

File: reviewers/pragmatist.md

Focus: Practical real-world impact.

Questions:

• "Your method improves accuracy 1% but adds 10x inference cost. Why use this?"
• "Have you tested on actual noisy real-world data?"
• "What happens on unseen input distributions?"
• "Is the complexity justified?"

Rates: practical_impact, deployment_feasibility, cost_benefit, real_world_validation

---

The Skeptic

File: reviewers/skeptic.md

Focus: Default position — every result is wrong until proven otherwise.

Questions:

• "Show me the raw data, not averaged plots."
• "Your error bars are suspiciously small. Explain."
• "Your p-value is 0.049 with n=12. I've seen this before."
• "Explain the result in your appendix you don't mention in the main text."

Rates: evidence_quality, robustness, transparency, honesty_in_reporting

---

The Historian

File: reviewers/historian.md

Focus: Prior art and citation accuracy.

Questions:

• "This result was published in 1972. You didn't cite it."
• "Your 'novel' method is a minor modification of Smith 1998."
• "The citation you give for this claim is wrong."
• "This problem formulation was solved 30 years ago."

Rates: citation_accuracy, prior_art_coverage, historical_context, novelty_accuracy

---

The Methodologist

File: reviewers/methodologist.md

Focus: Statistical and methodological correctness.

Questions:

• "You used a t-test but you have 4 groups. Why not ANOVA?"
• "Your sample size is 12 with effect size 0.8. Power analysis says n=30."
• "Did you correct for multiple comparisons?"
• "Your data is bimodal but you're using a normality-assuming test."

Rates: statistical_correctness, experimental_design, power_analysis, methodological_rigor

---

The Ethicist

File: reviewers/ethicist.md

Focus: Societal impact and fairness.

Questions:

• "Who is harmed by this technology?"
• "Your training data — was it ethically sourced?"
• "Your model performs differently across demographics. Did you measure this?"
• "Where's your dual-use discussion?"

Rates: ethical_consideration, fairness_analysis, dual_use_awareness, data_ethics

---

The Competitor

File: reviewers/competitor.md

Focus: Deep knowledge of the closest competing work.

Questions:

• "Your comparison uses suboptimal hyperparameters. Here are the correct ones."
• "This contribution is minor. The core idea is from our 2023 paper."
• "Your method is orthogonal to ours — they could be combined. You didn't explore this."

Rates: incremental_novelty, baseline_fairness, comparison_honesty, competitive_awareness

---

The Student

File: reviewers/student.md

Focus: Clarity and accessibility for newcomers.

Questions:

• "What does [term] mean? You defined it 10 pages ago."
• "Figure 3 makes no sense to me."
• "The paper assumes I know [concept]. I don't."
• "Why should I care? The introduction doesn't say."

Rates: clarity, exposition, accessibility, self_containedness

---

The Dreamer

File: reviewers/dreamer.md

Focus: Untapped potential and moonshots.

Questions:

• "What if you scaled this 100x?"
• "Your method solves X. What if the same approach solves Y, Z, W?"
• "What would it take to deploy this in the real world?"
• "If you had unlimited compute, what would you do differently?"

Rates: ambition, future_potential, breadth_of_impact, moonshot_thinking