add dreamgym to index

Author: bhavsarpratik

research-papers/index.html

@@ -70,6 +70,14 @@ <h1 class="heading-2">Research Papers</h1>
                                 </td>
                                 <td class="description">Synthesis of Toby Ord's half-life framework with METR's exponential growth analysis. Reveals AI agents fail at constant rate per minute (half-life model) while capabilities double every 7 months. Projects specific reliability thresholds: 90% reliability requires 1/7 task duration reduction, current models achieve 50-minute tasks at 50% success. Predicts month-long task automation by 2030, with practical architecture patterns for current reliability levels.</td>
                             </tr>
+                            <tr>
+                                <td>
+                                    <a href="dreamgym_report.html" target="_blank" class="paper-link">
+                                        <strong>DreamGym: Scaling Agent Learning via Experience Synthesis</strong>
+                                    </a>
+                                </td>
+                                <td class="description">Breakthrough framework for training AI agents through synthetic experience synthesis. Introduces reasoning-based experience model that simulates environment dynamics, enabling scalable reinforcement learning without costly real-world interactions. Achieves 30%+ improvement on non-RL-ready tasks like WebArena using zero real environment interactions, while matching state-of-the-art on traditional benchmarks. Addresses four critical challenges: costly rollouts, limited task diversity, unreliable rewards, and infrastructure complexity.</td>
+                            </tr>
                             <tr>
                                 <td>
                                     <a href="deepseek_ocr_report.html" target="_blank" class="paper-link">