Structured Context Engineering for File-Native Agentic Systems: Evaluating Schema Accuracy, Format Effectiveness, and Multi-File Navigation at Scale

Large Language Model agents increasingly operate external systems through programmatic interfaces, yet practitioners lack empirical guidance on how to structure the context these agents consume. Using SQL generation as a proxy for programmatic agent operations, we present a systematic study of context engineering for structured data, comprising 9,649 experiments across 11 models, 4 formats (YAML, Markdown, JSON, Token-Oriented Object Notation [TOON]), and schemas ranging from 10 to 10,000 tables. Our findings challenge common assumptions. First, architecture choice is model-dependent: file-based context retrieval improves accuracy for frontier-tier models (Claude, GPT, Gemini; +2.7%, p=0.029) but shows mixed results for open source models (aggregate -7.7%, p<0.001), with deficits varying substantially by model. Second, format does not significantly affect aggregate accuracy (chi-squared=2.45, p=0.484), though individual models, particularly open source, exhibit format-specific sensitivities. Third, model capability is the dominant factor, with a 21 percentage point accuracy gap between frontier and open source tiers that dwarfs any format or architecture effect. Fourth, file-native agents scale to 10,000 tables through domain-partitioned schemas while maintaining high navigation accuracy. Fifth, file size does not predict runtime efficiency: compact formats can consume significantly more tokens at scale due to format-unfamiliar search patterns. These findings provide practitioners with evidence-based guidance for deploying LLM agents on structured systems, demonstrating that architectural decisions should be tailored to model capability rather than assuming universal best practices.

Certifiable Boolean Reasoning Is Universal

The proliferation of agentic systems has thrust the reasoning capabilities of AI into the forefront of contemporary machine learning. While it is known that there \emph{exist} neural networks which can reason through any Boolean task $f:\{0,1\}^B\to\{0,1\}$, in the sense that they emulate Boolean circuits with fan-in $2$ and fan-out $1$ gates, trained models have been repeatedly demonstrated to fall short of these theoretical ideals. This raises the question: \textit{Can one exhibit a deep learning model which \textbf{certifiably} always reasons and can \textbf{universally} reason through any Boolean task?} Moreover, such a model should ideally require few parameters to solve simple Boolean tasks. We answer this question affirmatively by exhibiting a deep learning architecture which parameterizes distributions over Boolean circuits with the guarantee that, for every parameter configuration, a sample is almost surely a valid Boolean circuit (and hence admits an intrinsic circuit-level certificate). We then prove a universality theorem: for any Boolean $f:\{0,1\}^B\to\{0,1\}$, there exists a parameter configuration under which the sampled circuit computes $f$ with arbitrarily high probability. When $f$ is an $\mathcal{O}(\log B)$-junta, the required number of parameters scales linearly with the input dimension $B$. Empirically, on a controlled truth-table completion benchmark aligned with our setting, the proposed architecture trains reliably and achieves high exact-match accuracy while preserving the predicted structure: every internal unit is Boolean-valued on $\{0,1\}^B$. Matched MLP baselines reach comparable accuracy, but only about $10\%$ of hidden units admit a Boolean representation; i.e.\ are two-valued over the Boolean cube.

Rethinking Weight Tying: Pseudo-Inverse Tying for Stable LM Training and Updates

Weight tying is widely used in compact language models to reduce parameters by sharing the token table between the input embedding and the output projection. However, weight sharing does not guarantee a stable token interface: during training, the correspondence between encoding tokens into hidden states and decoding hidden states into logits can drift, worsening optimization sensitivity and making post-training interventions such as editing, patching, and lightweight adaptation less predictable. We propose Pseudo-Inverse Tying (PIT), which synchronizes embedding and unembedding as coupled projections of a shared latent token memory, guaranteeing a pseudo-inverse-consistent interface throughout training. PIT maintains an orthonormal shared memory, obtained by thin polar decomposition for teacher initialization or random orthonormal initialization from scratch, and introduces a fully learned symmetric positive definite hidden-space transform parameterized via a Cholesky factor. The output head applies this transform to hidden states before the vocabulary projection, while the embedding applies the inverse transform to token vectors using stable triangular solves, avoiding explicit pseudo-inverse recomputation and any vocabulary-sized auxiliary parameters. We evaluate PIT on on-device models spanning 256M-1.3B parameters across pretraining and adaptation, and consistently observe improved training stability, stronger layerwise semantic consistency, and substantially reduced side effects.

An Anatomy-specific Guidewire Shaping Robot for Improved Vascular Navigation

Neuroendovascular access often relies on passive microwires that are hand-shaped at the back table and then used to track a microcatheter to the target. Neuroendovascular surgeons determine the shape of the wire by examining the patient pre-operative images and using their experience to identify anatomy specific shapes of the wire that would facilitate reaching the target. This procedure is particularly complex in convoluted anatomical structures and is heavily dependent on the level of expertise of the surgeon. Towards enabling standardized autonomous shaping, we present a bench-top guidewire shaping robot capable of producing navigation-specific desired wire configurations. We present a model that can map the desired wire shape into robot actions, calibrated using experimental data. We show that the robot can produce clinically common tip geometries (C, S, Angled, Hook) and validate them with respect to the model-predicted shapes in 2D. Our model predicts the shape with a Root Mean Square (RMS) error of 0.56mm across all shapes when compared to the experimental results. We also demonstrate 3D tip shaping capabilities and the ability to traverse complex endoluminal navigation from the petrous Internal Carotid Artery (ICA) to the Posterior Communicating Artery (PComm).

Prediction of Critical Heat Flux in Rod Bundles Using Tube-Based Hybrid Machine Learning Models in CTF

The prediction of critical heat flux (CHF) using machine learning (ML) approaches has become a highly active research activity in recent years, the goal of which is to build models more accurate than current conventional approaches such as empirical correlations or lookup tables (LUTs). Previous work developed and deployed tube-based pure and hybrid ML models in the CTF subchannel code, however, full-scale reactor core simulations require the use of rod bundle geometries. Unlike isolated subchannels, rod bundles experience complex thermal hydraulic phenomena such as channel crossflow, spacer grid losses, and effects from unheated conductors. This study investigates the generalization of ML-based CHF prediction models in rod bundles after being trained on tube-based CHF data. A purely data-driven DNN and two hybrid bias-correction models were implemented in the CTF subchannel code and used to predict CHF location and magnitude in the Combustion Engineering 5-by-5 bundle CHF test series. The W-3 correlation, Bowring correlation, and Groeneveld LUT were used as baseline comparators. On average, all three ML-based approaches produced magnitude and location predictions more accurate than the baseline models, with the hybrid LUT model exhibiting the most favorable performance metrics.

Grables: Tabular Learning Beyond Independent Rows

Tabular learning is still dominated by row-wise predictors that score each row independently, which fits i.i.d. benchmarks but fails on transactional, temporal, and relational tables where labels depend on other rows. We show that row-wise prediction rules out natural targets driven by global counts, overlaps, and relational patterns. To make "using structure" precise across architectures, we introduce grables: a modular interface that separates how a table is lifted to a graph (constructor) from how predictions are computed on that graph (node predictor), pinpointing where expressive power comes from. Experiments on synthetic tasks, transaction data, and a RelBench clinical-trials dataset confirm the predicted separations: message passing captures inter-row dependencies that row-local models miss, and hybrid approaches that explicitly extract inter-row structure and feed it to strong tabular learners yield consistent gains.

Failure is Feedback: History-Aware Backtracking for Agentic Traversal in Multimodal Graphs

Open-domain multimodal document retrieval aims to retrieve specific components (paragraphs, tables, or images) from large and interconnected document corpora. Existing graph-based retrieval approaches typically rely on a uniform similarity metric that overlooks hop-specific semantics, and their rigid pre-defined plans hinder dynamic error correction. These limitations suggest that a retriever should adapt its reasoning to the evolving context and recover intelligently from dead ends. To address these needs, we propose Failure is Feedback (FiF), which casts subgraph retrieval as a sequential decision process and introduces two key innovations. (i) We introduce a history-aware backtracking mechanism; unlike standard backtracking that simply reverts the state, our approach piggybacks on the context of failed traversals, leveraging insights from previous failures. (ii) We implement an economically-rational agentic workflow. Unlike conventional agents with static strategies, our orchestrator employs a cost-aware traversal method to dynamically manage the trade-off between retrieval accuracy and inference costs, escalating to intensive LLM-based reasoning only when the prior failure justifies the additional computational investment. Extensive experiments show that FiF achieves state-of-the-art retrieval on the benchmarks of MultimodalQA, MMCoQA and WebQA.

PluRel: Synthetic Data unlocks Scaling Laws for Relational Foundation Models

Relational Foundation Models (RFMs) facilitate data-driven decision-making by learning from complex multi-table databases. However, the diverse relational databases needed to train such models are rarely public due to privacy constraints. While there are methods to generate synthetic tabular data of arbitrary size, incorporating schema structure and primary--foreign key connectivity for multi-table generation remains challenging. Here we introduce PluRel, a framework to synthesize multi-tabular relational databases from scratch. In a step-by-step fashion, PluRel models (1) schemas with directed graphs, (2) inter-table primary-foreign key connectivity with bipartite graphs, and, (3) feature distributions in tables via conditional causal mechanisms. The design space across these stages supports the synthesis of a wide range of diverse databases, while being computationally lightweight. Using PluRel, we observe for the first time that (1) RFM pretraining loss exhibits power-law scaling with the number of synthetic databases and total pretraining tokens, (2) scaling the number of synthetic databases improves generalization to real databases, and (3) synthetic pretraining yields strong base models for continued pretraining on real databases. Overall, our framework and results position synthetic data scaling as a promising paradigm for RFMs.

MeKi: Memory-based Expert Knowledge Injection for Efficient LLM Scaling

Scaling Large Language Models (LLMs) typically relies on increasing the number of parameters or test-time computations to boost performance. However, these strategies are impractical for edge device deployment due to limited RAM and NPU resources. Despite hardware constraints, deploying performant LLM on edge devices such as smartphone remains crucial for user experience. To address this, we propose MeKi (Memory-based Expert Knowledge Injection), a novel system that scales LLM capacity via storage space rather than FLOPs. MeKi equips each Transformer layer with token-level memory experts that injects pre-stored semantic knowledge into the generation process. To bridge the gap between training capacity and inference efficiency, we employ a re-parameterization strategy to fold parameter matrices used during training into a compact static lookup table. By offloading the knowledge to ROM, MeKi decouples model capacity from computational cost, introducing zero inference latency overhead. Extensive experiments demonstrate that MeKi significantly outperforms dense LLM baselines with identical inference speed, validating the effectiveness of memory-based scaling paradigm for on-device LLMs. Project homepage is at https://github.com/ningding-o/MeKi.

RDT2: Exploring the Scaling Limit of UMI Data Towards Zero-Shot Cross-Embodiment Generalization

Vision-Language-Action (VLA) models hold promise for generalist robotics but currently struggle with data scarcity, architectural inefficiencies, and the inability to generalize across different hardware platforms. We introduce RDT2, a robotic foundation model built upon a 7B parameter VLM designed to enable zero-shot deployment on novel embodiments for open-vocabulary tasks. To achieve this, we collected one of the largest open-source robotic datasets--over 10,000 hours of demonstrations in diverse families--using an enhanced, embodiment-agnostic Universal Manipulation Interface (UMI). Our approach employs a novel three-stage training recipe that aligns discrete linguistic knowledge with continuous control via Residual Vector Quantization (RVQ), flow-matching, and distillation for real-time inference. Consequently, RDT2 becomes one of the first models that simultaneously zero-shot generalizes to unseen objects, scenes, instructions, and even robotic platforms. Besides, it outperforms state-of-the-art baselines in dexterous, long-horizon, and dynamic downstream tasks like playing table tennis. See https://rdt-robotics.github.io/rdt2/ for more information.