Adam Morris

Julien Abadji, Marah Abdin, Connor Adams et al.

We present Laguna M.1 and Laguna XS.2, two Mixture-of-Experts foundation models built for long-horizon, agentic coding: M.1 has $225.8$B total parameters ($23.4$B activated per token) and XS.2 has $33.4$B total ($3$B activated). Both models were trained from scratch end-to-end inside the same internal system that we refer to as our Model Factory: a tightly-integrated stack of versioned data, training, evaluation, and inference components that turn model development into an industrial process. We describe the principles and design choices of the Model Factory and also detail the end-to-end training process of our models, throughout pre-training data and architecture, post-training stages, evaluation, and quantization. On agentic software engineering and terminal benchmarks (SWE-bench Verified, SWE-bench Multilingual, SWE-Bench Pro, and Terminal-Bench 2.0) M.1 and XS.2 are competitive with state-of-the-art open models in their respective weight classes. Laguna XS.2 weights are released under Apache~2.0 at https://huggingface.co/collections/poolside/laguna-xs2.

Adam Morris, Timothy Pelham, Edmund R. Hunt

This paper introduces a method for designing spatially intelligent robot swarm behaviors to localize concealed radio emitters. We use differential evolution to generate geometric patrol routes that localize unknown signals independently of emitter parameters, a key challenge in electromagnetic surveillance. Patrol shape and antenna type are shown to influence information gain, which in turn determines the effective triangulation coverage. We simulate a four-robot swarm across eight configurations, assigning pre-generated patrol routes based on a specified patrol shape and sensing capability (antenna type: omnidirectional or directional). An emitter is placed within the map for each trial, with randomized position, transmission power and frequency. Results show that omnidirectional localization success rates are driven primarily by source location rather than signal properties, with failures occurring most often when sources are placed in peripheral areas of the map. Directional antennas are able to overcome this limitation due to their higher gain and directivity, with an average detection success rate of 98.75% compared to 80.25% for omnidirectional. Average localization errors range from 1.01-1.30 m for directional sensing and 1.67-1.90 m for omnidirectional sensing; while directional sensing also benefits from shorter patrol edges. These results demonstrate that a swarm's ability to predict electromagnetic phenomena is directly dependent on its physical interaction with the environment. Consequently, spatial intelligence, realized here through optimized patrol routes and antenna selection, is a critical design consideration for effective robotic surveillance.

Dillon Plunkett, Adam Morris, Keerthi Reddy et al.

We have only limited understanding of how and why large language models (LLMs) respond in the ways that they do. Their neural networks have proven challenging to interpret, and we are only beginning to tease out the function of individual neurons and circuits within them. However, another path to understanding these systems is to investigate and develop their capacity to explain their own functioning. Here, we show that i) LLMs can accurately describe quantitative features of their own internal processes during certain kinds of decision-making and ii) that it is possible to improve these capabilities through training. To do so, we fine-tuned GPT-4o and GPT-4o-mini to make decisions in a wide variety of complex contexts (e.g., choosing between condos, loans, vacations, etc.) according to randomly-generated, quantitative preferences about how to weigh different attributes (e.g., the relative importance of natural light versus quiet surroundings for condos). We demonstrate that the LLMs can accurately report these preferences (i.e., the weights that they learned to give to different attributes during decision-making). Next, we demonstrate that these LLMs can be fine-tuned to explain their decision-making even more accurately. Finally, we demonstrate that this training generalizes: It improves the ability of the models to accurately explain how they make other complex decisions, not just decisions they have been fine-tuned to make. This work is a step towards training LLMs to accurately and broadly report on their own internal processes -- a possibility that would yield substantial benefits for interpretability, control, and safety.