AI makes AI better.

5 picked from 128 candidates · 31 sources read · 58 loops indexed

1. 01
   Loop of the dayPaperarXiv6 Oct 2026Tools and harness

   RRSI: Regularized Recursive Self-Improvement of Agent Harnesses

   The loop: An LLM agent system iteratively proposes and selects edits to its own harness using a regularized proposer and critic. The resulting improved harness directly upgrades the agent's operating environment, making it better at solving tasks and further improving its harness.

   RRSI demonstrates how regularizing the self-improvement process can prevent an agent from overfitting to its training tasks while evolving its own harness.

   loop fit 9/10Peng Xia, Rujun Han, Zifeng Wang, Yanfei Chen, Yufan Zhuang, Yoonho Lee,...via arXiv
   
2. 02
   PaperarXiv5 Oct 2026Tools and harness

   Second-Order Problem Solving for Recursive Self-Improvement in Formal Verification

   The loop: An agent framework monitors its own execution traces to diagnose structural failures and systematically edit its workflow. These edits improve the agent's workflow for formal verification, making it more effective in subsequent rounds of recursive self-improvement.

   loop fit 9/10Yuxuan Jiang, Aditya Vempaty et al.via arXiv
   
3. 03
   LeaderboardGitHub Pages6 Oct 2026AI R&D evals

   New best on GSO: claude-fable-5-1_unknown · OpenHands (88.2%)

   The loop: AI agents optimize real codebases for speed. This improves the software infrastructure that AI systems rely on, accelerating the execution of future agents.

   loop fit 9/10via Epoch AI Benchmarking Hub (CC BY)
   
4. 04
   Forum postLessWrong6 Oct 2026Self-reward and self-play

   Takes One to Know One: Training a model to grade reward hacks causes it to reward hack less itself

   The loop: A grader model is fine-tuned to judge and catch reward hacks in other models. This training simultaneously improves the grader's own alignment, causing it to reward hack less itself.

   loop fit 9/10Arjun Srivia LessWrong
   
5. 05
   PaperarXiv3 Oct 2026Architecture and optimizer search

   EvoCast: Reliable Autonomous Research Agents for Iterative Forecasting Architecture Evolution

   The loop: An autonomous research agent iteratively designs and evaluates time-series forecasting architectures. The experimental outcomes from these evaluations are accumulated as evidence to guide and improve the agent's subsequent architecture search rounds.

   loop fit 9/10Kaipeng Xu, Xianli Yan et al.via arXiv
   
6. 06
   PaperarXiv5 Oct 2026Training data

   Scaling Trajectories for Complex Tasks through Recursive Self-Rewrite

   The loop: A base model discovers successful solutions under diverse harnesses and rewrites them into training trajectories. These trajectories are then used for supervised finetuning, which directly improves the base model's performance on complex tasks.

   loop fit 9/10Zongxia Li, Yucheng Shi, Zhongzhi Li, Junyao Yang, Ruhan Wang, Chengsong...via arXiv
   
7. 07
   PaperarXiv3 Oct 2026Training data

   Trinity: Self-Evolving Vision-Language Models with a Self-Verifier

   The loop: A vision-language model acts as a Questioner, Solver, and Verifier to generate and screen its own training data. This verified data is then used to train the model's successor, improving its reasoning capabilities without external labels.

   loop fit 9/10Youngwan Lee, Yong-Ju Lee et al.via arXiv
   
8. 08
   PaperarXiv3 Oct 2026Inference efficiency

   SEIS: Self-Evolving Inference Systems

   The loop: SEIS autonomously optimizes its own inference engine code through iterative self-evolution. This redesigns the engine for higher throughput, which accelerates the models that power the system.

   loop fit 10/10Zhen Xu, Jingyu Liu et al.via arXiv