The integrated luminosity from the features of the polycyclic aromatic hydrocarbons (PAHs) exceeds the luminosity from atomic and molecular emission lines in the star-forming regions in galaxies and is a potential tracer of galaxy-scale star formation and molecular gas content of the high-redshift universe. We simulate the observable PAH spectra using the PRobe far-Infrared Mission for Astrophysics far-infrared enhanced survey spectrometer (FIRESS) and investigate the capability of the FIRESS low-resolution spectroscopy for observing PAH emission spectrum from high-redshift galaxies. Our investigation suggests that (1) PRIMA observations of PAH emission are $\gtrsim10$ times more efficient at detecting galaxies than the VLA observations of CO(1-0) for galaxies with the same infrared luminosity, (2) PRIMA/FIRESS can detect the PAH emission from galaxies with $L_{IR}\sim10^{12}L_{\odot}$ up to the end of reionization (and possibly beyond, if $L_{IR}\sim10^{13}L_{\odot}$), (3) the PAH band ratios measured from a full spectral fitting and from a simple flux "clipping" method are different and vary depending on the interstellar radiation field strength, and (4) PRIMA/FIRESS can also be used as the PAH mapping instrument to measure star formation and redshift of the galaxies in high-redshift protoclusters.

In recent research advancements within the community, large language models (LLMs) have sparked great interest in creating autonomous agents. However, current prompt-based agents often heavily rely on large-scale LLMs. Meanwhile, although fine-tuning methods significantly enhance the capabilities of smaller LLMs, the fine-tuned agents often lack the potential for self-reflection and self-improvement. To address these challenges, we introduce a novel agent framework named RetroAct, which is a framework that jointly optimizes both task-planning and self-reflective evolution capabilities in language agents. Specifically, we develop a two-stage joint optimization process that integrates imitation learning and reinforcement learning, and design an off-policy joint policy gradient optimization algorithm with imitation learning regularization to enhance the data efficiency and training stability in agent tasks. RetroAct significantly improves the performance of open-source models, reduces dependency on closed-source LLMs, and enables fine-tuned agents to learn and evolve continuously. We conduct extensive experiments across various testing environments, demonstrating RetroAct has substantial improvements in task performance and decision-making processes.