Research map / Optimization / Nonconvex and smooth optimization
Strongly-Typed Agents are Guaranteed to Interact Safely
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Use gradient descent in a safe convex game with smooth losses | Safe convex games · Safe convex games | Nash equilibrium convergence | Helped |
| Impose simultaneous factorization, commuting projections, and monotonic covariation | Gradient-based agents in strongly-typed games · Strongly-typed games | Safety | Helped |
| Use Newton's method on a strictly convex loss | Single-player open game · Strictly convex loss | Safety | Helped |
| Use feedback-alignment error signals satisfying the pseudoinverse condition | Neural network error propagation · Feedback alignment | Safety | Helped |
| Use kickback's truncated error signals | Neural network error propagation · Kickback | Safety | Helped |
| Use gradient descent in a zero-sum bilinear game | Two-player block game · Zero-sum bilinear game | Nash equilibrium convergence | Hurt |
| Use strong-typing rather than requiring a weighted potential game | Game safety analysis · Strongly-typed games | Safety | Helped |
| Treat potential-game structure alone as a safety guarantee | Potential games · Potential game example | Safety | Hurt |
Closest papers by meaning
- Robust and Diverse Multi-Agent Learning via Rational Policy Gradient (NeurIPS 2025 · similarity 0.86)
- Exploiting Structure in Offline Multi-Agent RL: The Benefits of Low Interaction Rank (ICLR 2025 · similarity 0.85)
- Protocols for Verifying Smooth Strategies in Bandits and Games (NeurIPS 2025 · similarity 0.85)
- Neural Interactive Proofs (ICLR 2025 · similarity 0.85)
- A Policy-Gradient Approach to Solving Imperfect-Information Games with Best-Iterate Convergence (ICLR 2025 · similarity 0.85)
- Breaking the Barrier: Enhanced Utility and Robustness in Smoothed DRL Agents (ICML 2024 · similarity 0.85)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.