Fork-and-Flush: Escaping Idea Basins in Autoresearch Agents
cs.LG
Submitted: 2026-10-05
Updated: 2026-10-05
Code: https://github.com/anthropics/original_performance_takehome
Sources
- Large Language Monkeys: Scaling Inference Compute with Repeated Sampling
- Training AI Scientists to Replicate Research
- Inverse Scaling in Test-Time Compute
- DiscoGen: Procedural Generation of Algorithm Discovery Tasks in Machine Learning
- Accelerating scientific discovery with Co-Scientist
- Population Based Training of Neural Networks
- Towards Robust Agentic CUDA Kernel Benchmarking, Verification, and Optimization
- The AI Scientist: Towards Fully Automated Open-Ended Scientific Discovery
- FrontierCS: Evolving Challenges for Evolving Intelligence
- DeepSeek-R1 Thoughtology: Let's think about LLM Reasoning
- Terminal-Bench: Benchmarking Agents on Hard, Realistic Tasks in Command Line Interfaces
- AlphaEvolve: A coding agent for scientific and algorithmic discovery
- West-of-N: Synthetic Preferences for Self-Improving Reward Models
- Wait, Wait, Wait... Why Do Reasoning Models Loop?
- Recursive Introspection: Teaching Language Model Agents How to Self-Improve
- Agent Laboratory: Using LLM Agents as Research Assistants
- Reflexion: Language Agents with Verbal Reinforcement Learning
- Beyond Human Data: Scaling Self-Training for Problem-Solving with Language Models
- Scaling LLM Test-Time Compute Optimally can be More Effective than Scaling Model Parameters
- Voyager: An Open-Ended Embodied Agent with Large Language Models
Related papers
- Polynomial-Augmented Neural Networks (PANNs) with Weak Orthogonality Constraints for Enhanced Function and PDE Approximation
- AIRL-S: Unifying Reinforcement Learning and Search-Based Test-Time Scaling via Adversarial Inverse Reinforcement Learning
- Transformers as Bayesian In-Context Experimenters: Smoothness-Adaptive Efficient ATE Estimation
- Convergence issues in Relational Concept Analysis based on AOC-posets
- Beliefs Beyond Posteriors: The Predictive Cost of ELBO Binding
- Understanding Diffusion Models via Ratio-Based Function Approximation with SignReLU Networks