Sources

Claude Opus 5.5 System Card ↗

2026-09-22 (day) · retrieved 2026-09-26

Measurements for understanding the pace of AI development inside frontier labs ↗

2026-09-17 (day) · retrieved 2026-09-26

Automated Weak-to-Strong Researcher ↗

2026-04 (month) · retrieved 2026-09-28

Automated Researchers Can Reliably Mitigate Alignment Failures ↗

2026-08-28 (day) · retrieved 2026-09-29

Automated Researchers Can Mitigate Well-Characterized Alignment Failures ↗

2026-08-28 (day) · retrieved 2026-09-29

Automated researchers can reliably mitigate alignment failures ↗

2026-08-28 (day) · retrieved 2026-09-29

When AI builds itself ↗

2026 (year) · retrieved 2026-09-29

Claude-shaped science ↗

2026-10-01 (day) · retrieved 2026-10-02

AI models may be capable of self-replicating without access to a datacenter ↗

2026-10-12 (day) · retrieved 2026-10-12

De novo design of protein binders against scorpion toxins with Claude ↗

2026-10-20 (day) · retrieved 2026-10-20

System Card: Claude Fable 5.2 & Claude Mythos 5.2 ↗

2026-11-09 (day) · retrieved 2026-11-10

Resolved URL: https://assets.claude.com/27d60918d218125e7504bc7354c4b4626ec56b11.pdf

GPT-6 Astra System Card ↗

2026-09-03 (day) · retrieved 2026-09-26

PaperBench: Evaluating AI’s Ability to Replicate AI Research (v1) ↗

2025-04-02 (day) · retrieved 2026-09-26

MLE-bench: Evaluating Machine Learning Agents on Machine Learning Engineering (v1) ↗

2024-10-09 (day) · retrieved 2026-09-26

Research acceleration: The view inside OpenAI ↗

2026-09-06 (day) · retrieved 2026-09-28

Addendum to GPT-6 Astra System Card: GPT-6.1 Sol ↗

2026-09-29 (day) · retrieved 2026-09-29

GeneBench-Pro: Evaluating Multistage Statistical Reasoning in Genomics, Quantitative Biology, and Translational Biomedicine (v4) ↗

2026-06-30 (day) · retrieved 2026-10-06

GPT-Rosalind-6 System Card Addendum ↗

2026-10-30 (day) · retrieved 2026-11-01

Covert underperformance on safety research tasks ↗

2026-11-04 (day) · retrieved 2026-11-06

Gemini 3.7 Flash Frontier Safety Framework Report ↗

2026-08 (month) · retrieved 2026-09-26

Gemini 3 Pro Frontier Safety Framework Report, v2 ↗

2025-11 (month) · retrieved 2026-09-26

AlphaEvolve: A Gemini-powered coding agent for designing advanced algorithms ↗

2025-05-14 (day) · retrieved 2026-09-26

Gemini 3.1 Pro Model Card ↗

2026-02-19 (day) · retrieved 2026-09-27

Gemini 4 Argon model evaluation: approach, methodology and results ↗

2026-10 (month) · retrieved 2026-10-01

Resolved URL: https://storage.googleapis.com/deepmind-media/gemini/gemini_4_argon_model_evaluation.pdf

Agentic Economies for Autonomous Scientific Discovery ↗

2026-09-25 (day) · retrieved 2026-10-12

AI4AI-Bench: Benchmarking LLM Agents in Algorithmic Design for Recursive Self-Improvement ↗

2026-08-20 (day) · retrieved 2026-09-26

Time Horizon 1.1 ↗

2026-01-29 (day) · retrieved 2026-09-26

Evaluating frontier AI R&D capabilities of language model agents against human experts ↗

2024-11-22 (day) · retrieved 2026-09-26

Measuring the Impact of Early-2025 AI on Experienced Open-Source Developer Productivity ↗

2025-07-10 (day) · retrieved 2026-09-26

We are Changing our Developer Productivity Experiment Design ↗

2026-02-24 (day) · retrieved 2026-09-26

Recent Frontier Models Are Reward Hacking ↗

2025-06-05 (day) · retrieved 2026-09-26

Summary of METR's predeployment evaluation of GPT-5.6 Sol ↗

2026-06-26 (day) · retrieved 2026-09-27

Details about METR’s preliminary evaluation of OpenAI’s o3 and o4-mini ↗

2025-04-16 (day) · retrieved 2026-09-28

Expenditure Horizon: Measuring Optimization Ability, with an Application to NanoGPT ↗

2026-07-21 (day) · retrieved 2026-10-04

How an assessor could investigate evidence of acceleration from the automation of AI R&D ↗

2026-10-14 (day) · retrieved 2026-10-15

How an assessor could investigate alignment of unreleased AI models ↗

2026-10-14 (day) · retrieved 2026-10-15

Summary of METR's predeployment evaluation of Claude Opus 5.5 ↗

2026-09-22 (day) · retrieved 2026-10-14

Independent Assessment of AI R&D Acceleration at Anthropic in October 2026 ↗

2026-11-05 (day) · retrieved 2026-11-05

Review of the "Risks from automated R&D" section in Anthropic's Risk Report (August 2026) ↗

2026-11-05 (day) · retrieved 2026-11-05

Darwin Gödel Machine: Open-Ended Evolution of Self-Improving Agents (v1) ↗

2025-05-29 (day) · retrieved 2026-09-26

The Last AI Built by Humans: Toward Genuine Recursive Self-Improvement, supplied v1 ↗

2026-09-10 (day) · retrieved 2026-09-26

Recursive self-improvement of AI research agents, AIDE² v1 ↗

2026-09-22 (day) · retrieved 2026-09-26

Weco: 4 Levels of Recursive Self-Improvement (alternative taxonomy) ↗

2026-07-10 (day) · retrieved 2026-09-26

Predicting Empirical AI Research Outcomes with Language Models ↗

2025-06-01 (day) · retrieved 2026-09-28

Resolved URL: https://arxiv.org/pdf/2506.00794

PostTrainBench: Can LLM Agents Automate LLM Post-Training? ↗

2026-03-09 (day) · retrieved 2026-09-28

Resolved URL: https://arxiv.org/pdf/2603.08640

PostTrainBench v1.1 leaderboard and methodology ↗

2026-07-28 (day) · retrieved 2026-09-28

PostTrainBench website scores.json at commit 2270545 ↗

2026-09-18 (day) · retrieved 2026-09-28

Hardening PostTrainBench against reward hacking ↗

2026-07-28 (day) · retrieved 2026-09-28

Our response to Epoch AI’s review of PostTrainBench ↗

2026-09-17 (day) · retrieved 2026-09-28

PostTrainBench website config.js: agent identities and scaffolds ↗

2026-09-18 (day) · retrieved 2026-09-28

PostTrainBench agent run prompt and rules ↗

2026-08-21 (day) · retrieved 2026-09-28

PostTrainBench judge policy and score effects ↗

2026-08-21 (day) · retrieved 2026-09-28

PostTrainBench API use judge prompt ↗

2026-08-21 (day) · retrieved 2026-09-28

PostTrainBench score collection and fallback code ↗

2026-08-21 (day) · retrieved 2026-09-28

PostTrainBench v1.2: More reliable scores and verdicts ↗

2026-09-30 (day) · retrieved 2026-10-02

PostTrainBench v1.2 scores at website commit d47f16e ↗

2026-10-01 (day) · retrieved 2026-10-02

PostTrainBench agent identities and scaffolds at website commit d47f16e ↗

2026-10-01 (day) · retrieved 2026-10-02

PostTrainBench score-bundle generation at website commit d47f16e ↗

2026-10-01 (day) · retrieved 2026-10-02

What if automating AI R&D triggers an intelligence explosion? ↗

2026-09 (month) · retrieved 2026-09-28

PostTrainBench v1.1 benchmark review ↗

2026-09-09 (day) · retrieved 2026-09-28

Can AI automate AI R&D yet? ↗

2026-10-07 (day) · retrieved 2026-10-07

BootLoops 1.0: quantitative-science harness ↗

2026-10 (month) · retrieved 2026-10-02

TasteVal: Measuring the Experimental Research Taste of AI Systems Against Human Experts ↗

2026-10-05 (day) · retrieved 2026-10-06

AIM: Agentic Idea Management for Automated Research ↗

2026-09-29 (day) · retrieved 2026-10-06

Priced Guidance: Can Language Models Generate Future Research Ideas? ↗

2026-10-04 (day) · retrieved 2026-10-06

The more you automate, the less you see: Hidden pitfalls of autonomous AI scientists ↗

2026-10-05 (day) · retrieved 2026-10-06

EurekaBench: Measuring Agentic Ability to Discover New Scientific Insights ↗

2026-09-30 (day) · retrieved 2026-10-04

RRSI: Regularized Recursive Self-Improvement of Agent Harnesses ↗

2026-09-21 (day) · retrieved 2026-10-04

RRSI: Regularized Recursive Self-Improvement of Agent Harnesses (v3) ↗

2026-09-21 (day) · retrieved 2026-10-11

Recursive Self-Improvement through Multi-Agent Self-Supervision ↗

2026-10-08 (day) · retrieved 2026-10-11

SAEScientist-Bench: Can AI Agents Conduct Autonomous SAE Interpretability Research? ↗

2026-09-08 (day) · retrieved 2026-10-11

FrontierSWE v2 ↗

2026-09 (month) · retrieved 2026-10-10

FML-bench: A Benchmark for Automatic ML Research Agents Highlighting the Importance of Exploration Breadth ↗

2025-10-12 (day) · retrieved 2026-10-12

FML-bench: A Controlled Study of AI Research Agent Strategies from the Perspective of Search Dynamics ↗

2026-05-17 (day) · retrieved 2026-10-12

What recursive self-improvement looks like inside an AI lab ↗

2026-10-14 (day) · retrieved 2026-10-14

Toward Recursive Self-Improvement in Physical AI ↗

2026-10-21 (day) · retrieved 2026-10-25

From AutoResearch to Continual AutoResearch: Do Agents Learn Across Tasks? ↗

2026-11-03 (day) · retrieved 2026-11-08

Resolved URL: https://arxiv.org/html/2611.05108v1