update.

rl-introduction.bbl

-\begin{thebibliography}{134}
+\begin{thebibliography}{136}
 \providecommand{\natexlab}[1]{#1}
 \providecommand{\url}[1]{\texttt{#1}}
 \expandafter\ifx\csname urlstyle\endcsname\relax
@@ -58,6 +58,13 @@ Charlie Chen, Sebastian Borgeaud, Geoffrey Irving, Jean-Baptiste Lespiau, Lauren
 \newblock \emph{ArXiv preprint}, abs/2302.01318, 2023{\natexlab{b}}.
 \newblock URL \url{https://arxiv.org/abs/2302.01318}.
 
+\bibitem[Chen et~al.(2026)Chen, Wang, Wang, Xie, Li, Wang, and Xu]{chen-etal:failure-attribution}
+Jianming Chen, Yawen Wang, Junjie Wang, Xiaofei Xie, Shoubin Li, Qing Wang, and Fanjiang Xu.
+\newblock Where did it go wrong? capability-oriented failure attribution for vision-and-language navigation agents.
+\newblock In \emph{Findings of the Association for Computational Linguistics: ACL 2026}, pp.\  28132--28150, San Diego, California, United States, July 2026. Association for Computational Linguistics.
+\newblock \doi{10.18653/v1/2026.findings-acl.1402}.
+\newblock URL \url{https://aclanthology.org/2026.findings-acl.1402/}.
+
 \bibitem[Chen et~al.(2024)Chen, Li, Yan, Wang, Gunaratna, Yadav, Tang, Srinivasan, Zhou, Huang, and Jin]{chen-etal:2023alpagasus}
 Lichang Chen, Shiyang Li, Jun Yan, Hai Wang, Kalpa Gunaratna, Vikas Yadav, Zheng Tang, Vijay Srinivasan, Tianyi Zhou, Heng Huang, and Hongxia Jin.
 \newblock Alpagasus: Training a better alpaca with fewer data.
@@ -450,6 +457,12 @@ Mingjie Sun, Zhuang Liu, Anna Bair, and J.~Zico Kolter.
 \newblock In \emph{The Twelfth International Conference on Learning Representations, {ICLR} 2024, Vienna, Austria, May 7-11, 2024}. OpenReview.net, 2024.
 \newblock URL \url{https://openreview.net/forum?id=PxoFut3dWW}.
 
+\bibitem[Sun et~al.(2026)Sun, Wang, Schmidt, Yeung-Levy, and Zhang]{sun-etal:learning-from-failure}
+Xueqiao Sun, Xiaohan Wang, Ludwig Schmidt, Serena Yeung-Levy, and Yuhui Zhang.
+\newblock Learning from failure: Inference-time self-improvement for computer-use agents.
+\newblock \emph{arXiv preprint arXiv:2606.31270}, 2026.
+\newblock URL \url{https://arxiv.org/abs/2606.31270}.
+
 \bibitem[Sun et~al.(2023)Sun, Shen, Cao, Liu, Li, Shen, Gan, Gui, Wang, Yang, et~al.]{sun-etal:2023aligning}
 Zhiqing Sun, Sheng Shen, Shengcao Cao, Haotian Liu, Chunyuan Li, Yikang Shen, Chuang Gan, Liang-Yan Gui, Yu-Xiong Wang, Yiming Yang, et~al.
 \newblock Aligning large multimodal models with factually augmented rlhf.

rl-introduction.bib

@@ -2,6 +2,31 @@
 
 
 
+
+
+
+
+@inproceedings{chen-etal:failure-attribution,
+    title     = {Where Did It Go Wrong? Capability-Oriented Failure Attribution for Vision-and-Language Navigation Agents},
+    author    = {Chen, Jianming and Wang, Yawen and Wang, Junjie and Xie, Xiaofei and Li, Shoubin and Wang, Qing and Xu, Fanjiang},
+    booktitle = {Findings of the Association for Computational Linguistics: ACL 2026},
+    pages     = {28132--28150},
+    address   = {San Diego, California, United States},
+    publisher = {Association for Computational Linguistics},
+    month     = jul,
+    year      = {2026},
+    doi       = {10.18653/v1/2026.findings-acl.1402},
+    url       = {https://aclanthology.org/2026.findings-acl.1402/}
+}
+
+@article{sun-etal:learning-from-failure,
+    title   = {Learning from Failure: Inference-Time Self-Improvement for Computer-Use Agents},
+    author  = {Sun, Xueqiao and Wang, Xiaohan and Schmidt, Ludwig and Yeung-Levy, Serena and Zhang, Yuhui},
+    journal = {arXiv preprint arXiv:2606.31270},
+    year    = {2026},
+    url     = {https://arxiv.org/abs/2606.31270}
+}
+
 @article{zhang-etal:autoenv,
   title         = {AutoEnv: Automated Environments for Measuring
                    Cross-Environment Agent Learning},

section6/section6.tex

@@ -280,7 +280,7 @@ This process is illustrated in Figure~\ref{fig:agent-environment-scaling}. It ty
 \item \textbf{Scenario Generation.} Based on the environment specification, we generate diverse initial states and user tasks. The tasks should cover different goals, constraints, and difficulty levels. For example, in a shopping environment, a simple task may ask the agent to purchase a single item, while a more difficult task may require it to compare multiple products.
 \item \textbf{Agent--Environment Interaction.} The agent interacts with the generated environment through multiple rounds of actions and observations. At each step, the agent selects a tool and provides the required arguments. The environment executes the action and returns an observation.
 \item \textbf{Outcome Verification.} After the interaction is completed, a verifier checks whether the task has been successfully solved. The verifier may inspect tool execution results, database states, or other structured records. 
-\item \textbf{Agent Learning.} Finally, the verified trajectories are used to improve the agent. Successful trajectories can serve as demonstrations for SFT, while task-level or step-level feedback can support RL training. Note that failed interactions are also useful because they reveal weaknesses in the current policy or gaps in the task distribution. For example, repeated failures on invalid API arguments may motivate the generation of additional scenarios involving parameter constraints. In this way, learning results can further guide scenario generation and form an iterative environment--agent improvement loop.
+\item \textbf{Agent Learning.} Finally, the verified trajectories are used to improve the agent. On the one hand, successful trajectories can serve as demonstrations for SFT, while task-level or step-level feedback can support RL training. On the other hand, failed interactions are also valuable because they reveal weaknesses in the current policy and gaps in the task distribution. This can guide the generation of targeted scenarios that exercise the agent's weak capabilities and provide more informative experience for subsequent training \citep{chen-etal:failure-attribution,sun-etal:learning-from-failure}.
 \end{itemize}
 
 
@@ -301,11 +301,24 @@ Scaling environments also introduces substantial heterogeneity. Different enviro
 
 \subsection{Learning from Agentic Experience}
 
+
+
+
+
 \subsubsection{Memory Management}
 
+
+
+
+
 \subsubsection{Skill Optimization}
 
 
+
+
+
+
+
 \subsubsection{Trajectory Refinement}