arXiv subjects

Andrew Chen

Publications and source records attributed to Andrew Chen.

At least 19 recordsLinked to original sources

arXiv2026

IRIS: Navigating and Reflecting on Writing Traces Using Intelligent Document Histories

Much of the text produced throughout the lifetime of a document is impermanent. In this paper, we explore how writing activity traces can be made visible and interactive to help writers navigate their document histories and understand their writing processes. Using the Flower and Hayes cognitive process model of writing, IRIS infers writing process states from keystroke logs and presents them using an AI-enhanced version history. IRIS provides three primary interactions: revision highlighting that shows local process histories in-situ, conceptual filters that constrain the version history by process type or topic, and natural language inquiry that lets writers pose reflective questions about their writing and process. Following a formative and a longitudinal study, we find that writers use the interfaces to locate specific revisions and understand the progression of their writing. They use system outputs as interpretive material, relating them to pre-existing beliefs and confirming, challenging, and deepening their understanding of their writing.

cs.HC↗

arXiv2026

Macaron-V1: Towards Open Continual Learning with Self-Improvement and Mixture-of-LoRA

Macaron-V1 is an open agent-model family for experiential intelligence: learning from experience in real environments and continuing to learn after deployment. It is organized around two system goals. Adaptation is pursued through recursive improvement of versioned model-harness pairs, where experience from one configuration is evaluated under an external contract and used to construct its successor. Collaboration is pursued via the Mixture-of-LoRA (MoL) architecture that freezes a base model, composes specialist LoRA adapters, and selects one LoRA per user turn. The flagship Macaron-V1-Venti (748B) combines a 744B GLM-5.2 base with four LoRAs for chat, agent, coding, and GenUI; the Qwen3.6-35B-based Macaron-V1-Tall (50B) uses the same design for local deployment. This report presents Macaron-V1 as a co-designed system spanning architecture, algorithms, and infrastructure. The MoL architecture supports continual learning through extensible LoRA specialists. The algorithm combines Model-Harness Co-design and recursive self-improvement loop, including the UI4A component-native GenUI harness, a stateful action substrate, versioned Harness Context Protocol contract, and the agentic RL framework MindForge. The supporting infrastructure includes the post-training platform MinT, the long-context RL method LongStraw, and stability techniques for sparse MoE and DSA base models. We evaluate Macaron-V1 on Personal Intelligence, GenUI, and general capability benchmarks against frontier baselines. Our results validate the current system, while compounding gains from continual learning and collective intelligence remain open questions.

cs.LG↗

arXiv2026

LongStraw: Long-Context RL Beyond 2M Tokens under a Fixed GPU Budget

Long-context RL post-training is constrained by the lifetime of state and gradients, not attention cost alone. In GRPO, one multi-million-token prompt must serve old-policy and reference scoring plus multiple policy responses, while conventional autograd keeps the prompt graph and all response graphs live alongside model weights, caches, and distributed communication buffers. We present LongStraw, an objective-aware, architecture-aware system for resident-state virtualization, response replay, and distributed-gradient execution. Its transaction captures the shared prompt without autograd, retains only the architecture-required state on explicitly owned pages, restores that state for each group member, scores old/reference branches without a graph, replays one policy response at a time with autograd, and accumulates the resulting gradients before one distributed finalization and optimizer step. This schedule bounds the live training graph by the response suffix while reusing the expensive prompt computation across the complete GRPO group. We instantiate this design for two incompatible model structures. Qwen3.6-27B combines 48 recurrent GDN layers with 16 full-attention layers; LongStraw keeps the compact recurrent state and physically CP8-sharded KV pages, composes global attention through cross-rank LSE/output merging, and performs blockwise response replay. GLM-5.2 combines a 78-layer MLA/DSA attention stack with a 256-expert, top-8 MoE tail. Its implementation keeps CP-sharded MLA latent pages and DSA indexer-key pages in CPU memory, stages one layer at a time, reconstructs IndexShare-aware global sparse selection over CP32, and dispatches routed response tokens over EP32. The two paths share one transaction contract while specializing the retained state, replay operator, and collective communication to the architecture...

cs.LG↗

arXiv2026

RoboTales: ROBOTic Anthropomorphic LEarning Systems

RoboTales is a low-cost robotic storytelling system that animates narratives using expressive sock puppetry. Implemented autonomously on a Baxter robot as a test case, RoboTales synchronizes narration, gestures, and mouth movements to perform character-driven stories. In a pilot study, puppet-based storytelling outperformed a gesture-only mode, producing higher HRIES ratings and improved story recall, suggesting that embodied puppetry enhances engagement and narrative comprehension. Designed to be modular and platform-agnostic, RoboTales can be adapted to other manipulators and offers a screen-free alternative to passive media, supporting future deployment in child-centered learning environments.

cs.RO↗

arXiv2026

An Irrelevance Theorem for Risk Aversion and Time-Varying Risk

We provide a theorem on the role of risk and risk attitudes in macroeconomic models that clarifies and extends the Tallarini (2000) separation result. Under (1) separation of intertemporal and risk preferences, (2) separation of drivers of first and higher moments in the model primitives, and (3) approximate linearity of constraints, risk aversion and time-varying risk are irrelevant for the elasticity of any endogenous variable with respect to state variables that don't drive variation in higher moments. We discuss how models generate a more prominent role for risk by ``breaking'' or ``adapting'' to the assumptions in the theorem.

econ.TH↗

arXiv2026

On the Scaling of PEFT: Towards Million Personal Models of Trillion Parameters

Parameter-efficient fine-tuning (PEFT) is usually treated as a cheaper alternative to full fine-tuning. We study a broader role: small trainable adapters as persistent local state on top of strong shared foundation models. In this framing, the base model provides shared competence while adapters carry instance-specific behavior such as preferences, skills, tool habits, and memory-like updates. We organize the problem around three scaling axes: Scale Up, where stronger shared priors make small local updates more useful; Scale Down, where we study how small adapters can be while remaining reliable; and Scale Out, where many persistent adapted instances coexist. MinT provides one infrastructure example for managing adapter identity, revision, provenance, evaluation, and serving residency. Together, the results suggest that PEFT can be a compact substrate for persistent personal models rather than only a budget substitute for full fine-tuning.

cs.LG↗

arXiv2026

Macaron-A2UI: A Model for Generative UI in Personal Agents

As personal agents evolve to handle complex, user-centric tasks, static plain-text chat is rapidly becoming a bottleneck. Generative UI emerges as the necessary new interface layer, dynamically synthesizing the right controls, options, and state from the interaction context in real time. We present Macaron-A2UI, a model for Generative UI in personal agents. Our goal is to move beyond text-only interaction by enabling agents to generate natural language together with lightweight, executable UI actions for information collection, preference refinement, confirmation, and multi-goal organization. We build a large-scale Generative UI corpus from heterogeneous dialogue sources, introduce A2UI-Bench for controlled evaluation, and train 30B, 235B and 754B models with parameter-efficient LoRA-based supervised fine-tuning followed by reward-driven reinforcement learning. The best Macaron-A2UI model reaches 75.6 overall on A2UI-Bench without explicit schema hints, surpassing the strongest full-schema frontier baseline. We release the models, benchmark, and evaluation protocol to support future work on Generative UI for personal agents.

cs.HC↗

arXiv2026

MinT: Managed Infrastructure for Training and Serving Millions of LLMs

We present MindLab Toolkit (MinT), a managed infrastructure system for Low-Rank Adaptation (LoRA) post-training and online serving. MinT targets a setting where many trained policies are produced over a small number of expensive base-model deployments. Instead of materializing each policy as a merged full checkpoint, MinT keeps the base model resident and moves exported LoRA adapter revisions through rollout, update, export, evaluation, serving, and rollback, hiding distributed training, serving, scheduling, and data movement behind a service interface. MinT scales this path along three axes. Scale Up extends LoRA RL to frontier-scale dense and MoE architectures, including MLA and DSA attention paths, with training and serving validated beyond 1T total parameters. Scale Down moves only the exported LoRA adapter, which can be under 1% of base-model size in rank-1 settings; adapter-only handoff reduces the measured step by 18.3x on a 4B dense model and 2.85x on a 30B MoE, while concurrent multi-policy GRPO shortens wall time by 1.77x and 1.45x without raising peak memory. Scale Out separates durable policy addressability from CPU/GPU working sets: a tensor-parallel deployment supports 10^6-scale addressable catalogs (measured single-engine sweeps through 100K) and thousand-adapter active waves at cluster scale, with cold loading treated as scheduled service work and packed MoE LoRA tensors improving live engine loading by 8.5-8.7x. MinT thus manages million-scale LoRA policy catalogs while training and serving selected adapter revisions over shared 1T-class base models.

cs.LG↗

arXiv2024

Clinical Applications of Plantar Pressure Measurement

Plantar pressure measurements can provide valuable insight into various health characteristics in patients. In this study, we describe different plantar pressure devices available on the market and their clinical relevance. Current devices are either platform-based or wearable and consist of a variety of sensor technologies: resistive, capacitive, piezoelectric, and optical. The measurements collected from any of these sensors can be utilized for a range of clinical applications including patients with diabetes, trauma, deformity and cerebral palsy, stroke, cervical myelopathy, ankle instability, sports injuries, and Parkinsons disease. However, the proper technology should be selected based on the clinical need and the type of tests being performed on the device. In this review we provide the reader with a simple overview of the existing technologies their advantages and disadvantages and provide application examples for each. Moreover, we suggest new areas in orthopaedic that plantar pressure mapping technology can be utilized for increased quality of care.

eess.SP↗

arXiv2023

ACER: An AST-based Call Graph Generator Framework

We introduce ACER, an AST-based call graph generator framework. ACER leverages tree-sitter to interface with any language. We opted to focus on generators that operate on abstract syntax trees (ASTs) due to their speed and simplicitly in certain scenarios; however, a fully quantified intermediate representation usually provides far better information at the cost of requiring compilation. To evaluate our framework, we created two context-insensitive Java generators and compared them to existing open-source Java generators.

cs.SE↗

arXiv2023

Gammapy: A Python package for gamma-ray astronomy

In this article, we present Gammapy, an open-source Python package for the analysis of astronomical $\gamma$-ray data, and illustrate the functionalities of its first long-term-support release, version 1.0. Built on the modern Python scientific ecosystem, Gammapy provides a uniform platform for reducing and modeling data from different $\gamma$-ray instruments for many analysis scenarios. Gammapy complies with several well-established data conventions in high-energy astrophysics, providing serialized data products that are interoperable with other software packages. Starting from event lists and instrument response functions, Gammapy provides functionalities to reduce these data by binning them in energy and sky coordinates. Several techniques for background estimation are implemented in the package to handle the residual hadronic background affecting $\gamma$-ray instruments. After the data are binned, the flux and morphology of one or more $\gamma$-ray sources can be estimated using Poisson maximum likelihood fitting and assuming a variety of spectral, temporal, and spatial models. Estimation of flux points, likelihood profiles, and light curves is also supported. After describing the structure of the package, we show, using publicly available $\gamma$-ray data, the capabilities of Gammapy in multiple traditional and novel $\gamma$-ray analysis scenarios, such as spectral and spectro-morphological modeling and estimations of a spectral energy distribution and a light curve. Its flexibility and power are displayed in a final multi-instrument example, where datasets from different instruments, at different stages of data reduction, are simultaneously fitted with an astrophysical flux model.

astro-ph.IM↗

arXiv2023

Paarl Africa Underground Laboratory (PAUL)

Establishing a deep underground physics laboratory to study, amongst others, double beta decay, geoneutrinos, reactor neutrinos and dark matter has been discussed for more than a decade within the austral African physicists' community. PAUL, the Paarl Africa Underground Laboratory, is an initiative foreseeing an open international laboratory devoted to the development of competitive science in the austral region. It has the advantage that the location, the Huguenot tunnel, exists already and the geology and the environment of the site is appropriate for an experimental facility. The paper describes the PAUL initiative, presents the physics prospects and discusses the capacity for building the future experimental facility.

hep-ex↗

arXiv2022

TeleSHift: Telexisting TUI for Physical Collaboration & Interaction

Current workflow on co-editing and simultaneous presentation of 3-D shapes is confined to on-screen manipulation, which causes loss of perceived information when presenting perceptual concepts or complex shapes between members. Thus, we create TeleSHift, a 3-D tangible user interface (TUI) with a telexisting communication framework for group-based collaboration and demonstration. In this work, we present a larger-scaled proof-of-concept prototype providing hands-on operation for shape-based interactions including multi-sided collaboration and one-to-many presentation. In contrast to previous works, we further extend the use of TUIs to support cooperative tasks with telexistence while enabling the linkage of manipulable bits to provide a better user experience and interactivity.

cs.HC↗

arXiv2022

Lightweight starshade position sensing with convolutional neural networks and simulation-based inference

Starshades are a leading technology to enable the direct detection and spectroscopic characterization of Earth-like exoplanets. To keep the starshade and telescope aligned over large separations, reliable sensing of the peak of the diffracted light of the occluded star is required. Current techniques rely on image matching or model fitting, both of which put substantial computational burdens on resource-limited spacecraft computers. We present a lightweight image processing method based on a convolutional neural network paired with a simulation-based inference technique to estimate the position of the spot of Arago and its uncertainty. The method achieves an accuracy of a few centimeters across the entire pupil plane, while only requiring 1.6 MB in stored data structures and 5.3 MFLOPs (million floating point operations) per image at test time. By deploying our method at the Princeton Starshade Testbed, we demonstrate that the neural network can be trained on simulated images and used on real images, and that it can successfully be integrated in the control system for closed-loop formation flying.

astro-ph.IM↗

arXiv2022

Scaling Up Models and Data with $\texttt{t5x}$ and $\texttt{seqio}$

Recent neural network-based language models have benefited greatly from scaling up the size of training datasets and the number of parameters in the models themselves. Scaling can be complicated due to various factors including the need to distribute computation on supercomputer clusters (e.g., TPUs), prevent bottlenecks when infeeding data, and ensure reproducible results. In this work, we present two software libraries that ease these issues: $\texttt{t5x}$ simplifies the process of building and training large language models at scale while maintaining ease of use, and $\texttt{seqio}$ provides a task-based API for simple creation of fast and reproducible training data and evaluation pipelines. These open-source libraries have been used to train models with hundreds of billions of parameters on datasets with multiple terabytes of training data. Along with the libraries, we release configurations and instructions for T5-like encoder-decoder models as well as GPT-like decoder-only architectures. $\texttt{t5x}$ and $\texttt{seqio}$ are open source and available at https://github.com/google-research/t5x and https://github.com/google/seqio, respectively.

cs.LG↗

arXiv2021

Deterministic Graph Coloring in the Streaming Model

Recent breakthroughs in graph streaming have led to the design of single-pass semi-streaming algorithms for various graph coloring problems such as $(\Delta+1)$-coloring, degeneracy-coloring, coloring triangle-free graphs, and others. These algorithms are all randomized in crucial ways and whether or not there is any deterministic analogue of them has remained an important open question in this line of work. We settle this fundamental question by proving that there is no deterministic single-pass semi-streaming algorithm that given a graph $G$ with maximum degree $\Delta$, can output a proper coloring of $G$ using any number of colors which is sub-exponential in $\Delta$. Our proof is based on analyzing the multi-party communication complexity of a related communication game, using random graph theory type arguments that may be of independent interest. We complement our lower bound by showing that just one extra pass over the input allows one to recover an $O(\Delta^2)$ coloring via a deterministic semi-streaming algorithm. This result is further extended to an $O(\Delta)$ coloring in $O(\log{\Delta})$ passes even in dynamic streams.

cs.DS↗

arXiv2021

Spatial Graph Attention and Curiosity-driven Policy for Antiviral Drug Discovery

We developed Distilled Graph Attention Policy Network (DGAPN), a reinforcement learning model to generate novel graph-structured chemical representations that optimize user-defined objectives by efficiently navigating a physically constrained domain. The framework is examined on the task of generating molecules that are designed to bind, noncovalently, to functional sites of SARS-CoV-2 proteins. We present a spatial Graph Attention (sGAT) mechanism that leverages self-attention over both node and edge attributes as well as encoding the spatial structure -- this capability is of considerable interest in synthetic biology and drug discovery. An attentional policy network is introduced to learn the decision rules for a dynamic, fragment-based chemical environment, and state-of-the-art policy gradient techniques are employed to train the network with stability. Exploration is driven by the stochasticity of the action space design and the innovation reward bonuses learned and proposed by random network distillation. In experiments, our framework achieved outstanding results compared to state-of-the-art algorithms, while reducing the complexity of paths to chemical synthesis.

cs.LG↗

arXiv2020

Event reconstruction for KM3NeT/ORCA using convolutional neural networks

The KM3NeT research infrastructure is currently under construction at two locations in the Mediterranean Sea. The KM3NeT/ORCA water-Cherenkov neutrino detector off the French coast will instrument several megatons of seawater with photosensors. Its main objective is the determination of the neutrino mass ordering. This work aims at demonstrating the general applicability of deep convolutional neural networks to neutrino telescopes, using simulated datasets for the KM3NeT/ORCA detector as an example. To this end, the networks are employed to achieve reconstruction and classification tasks that constitute an alternative to the analysis pipeline presented for KM3NeT/ORCA in the KM3NeT Letter of Intent. They are used to infer event reconstruction estimates for the energy, the direction, and the interaction point of incident neutrinos. The spatial distribution of Cherenkov light generated by charged particles induced in neutrino interactions is classified as shower- or track-like, and the main background processes associated with the detection of atmospheric neutrinos are recognized. Performance comparisons to machine-learning classification and maximum-likelihood reconstruction algorithms previously developed for KM3NeT/ORCA are provided. It is shown that this application of deep convolutional neural networks to simulated datasets for a large-volume neutrino telescope yields competitive reconstruction results and performance improvements with respect to classical approaches.

astro-ph.IM↗