RNA co-transcriptionality, where RNA is spliced or folded during transcription from DNA templates, offers promising potential for molecular programming. It enables programmable folding of nanoscale RNA structures and has recently been shown to be Turing universal. While post-transcriptional splicing is well studied, co-transcriptional splicing is gaining attention for its efficiency, though its unpredictability still remains a challenge. In this paper, we focus on engineering co-transcriptional splicing, not only as a natural phenomenon but as a programmable mechanism for generating specific RNA target sequences from DNA templates. The problem we address is whether we can encode a set of RNA sequences for a given system onto a DNA template word, ensuring that all the sequences are generated through co-transcriptional splicing. Given that finding the optimal encoding has been shown to be NP-complete under the various energy models considered [Da-Jung Cho et al., 2025], we propose a practical alternative approach under the logarithmic energy model. More specifically, we provide a construction that encodes an arbitrary nondeterministic finite automaton (NFA) into a circular DNA template from which co-transcriptional splicing produces all sequences accepted by the NFA. As all finite languages can be efficiently encoded as NFA, this framework solves the problem of finding small DNA templates for arbitrary target sets of RNA sequences. The quest to obtain the smallest possible such templates naturally leads us to consider the problem of minimizing NFAs and certain practically motivated variants of it, but as we show, those minimization problems are computationally intractable.

Sentence embedding models are typically trained using the Contrastive Learning (CL) method, which works by pulling similar semantics closer and pushing dissimilar ones away. Recent studies have shown that utilizing a multi-teacher ranking distillation approach, which assigns fine-grained rankings to sentences, enables the generation of smoother sentence similarity representations and results in higher-quality sentence embeddings. However, the effectiveness of distillation may be limited by the capacity of the student model. A simple student model with fewer parameters may struggle to approximate a highly complex teacher model, potentially leading to overfitting on certain datasets or specific aspects of the task. To address this, we propose MaCSE, a multi-agent ranking distillation framework that dynamically selects and optimizes teacher model contributions across training stages. MaCSE employs a Centralized Training with Decentralized Execution (CTDE) paradigm, enabling collaborative agent interactions to adaptively adjust teacher fusion weights based on training dynamics. Experimental results on Semantic Textual Similarity and transfer tasks demonstrate that MaCSE outperforms most existing baselines and even rivals methods using large language models for sentence representation. Our implementation is available at GitHub1.

Abstract Computing via synthetically engineered bacteria is a vibrant and active field with numerous applications in bio-production, bio-sensing, and medicine. Motivated by the lack of robustness and by resource limitation inside single cells, distributed approaches with communication among bacteria have recently gained in interest. In this paper, we focus on the problem of population growth happening concurrently, and possibly interfering, with the desired bio-computation. Specifically, we present a fast protocol in systems with continuous population growth for the majority consensus problem and prove that it correctly identifies the initial majority among two inputs with high probability if the initial difference is $$\varOmega (\sqrt{n\log n})$$ Ω ( n log n ) where n is the total initial population. We also present a fast protocol that correctly computes the Nand of two inputs with high probability. By combining Nand gates with the majority consensus protocol as an amplifier, it is possible to compute arbitrary Boolean functions. Finally, we extend the protocols to several biologically relevant settings. We simulate a plausible implementation of a noisy Nand gate with engineered bacteria. In the context of continuous cultures with a constant outflow and a constant inflow of fresh media, we demonstrate that majority consensus is achieved only if the flow is slower than the maximum growth rate. Simulations suggest that flow increases consensus time over a wide parameter range. The proposed protocols help set the stage for bio-engineered distributed computation that directly addresses continuous stochastic population growth.

Computing with synthetically engineered bacteria is a vibrant and active field with numerous applications in bio-production, bio-sensing, and medicine. Motivated by the lack of robustness and by resource limitation inside single cells, distributed approaches with communication among bacteria have recently gained in interest. In this paper, we focus on the problem of population growth happening concurrently, and possibly interfering, with the desired bio-computation. Specifically, we present a fast protocol in systems with continuous population growth for the majority consensus problem and prove that it correctly identifies the initial majority among two inputs with high probability if the initial difference is $Ω(\sqrt{n\log n})$ where $n$ is the total initial population. We also present a fast protocol that correctly computes the NAND of two inputs with high probability. We demonstrate that combining the NAND gate protocol with the continuous-growth majority consensus protocol, using the latter as an amplifier, it is possible to implement circuits computing arbitrary Boolean functions.