Ten simple rules for teaching data science
by Tiffany A. Timbers, Mine Çetinkaya-Rundel
Academic Intelligence · Curated Daily
AcademicHub 汇聚顶级期刊与预印本平台的实时文献。定制您的专属科研雷达,利用大语言模型自动生成交叉领域文献分析简报。
by Tiffany A. Timbers, Mine Çetinkaya-Rundel
arXiv:2606.13868v1 Announce Type: cross Abstract: We present an end-to-end pipeline for estimating stellar parameters from Sloan Digital Sky Survey Data Release 12 spectra using a fully connected multitask neural network with residual blocks, whose hyperparameters are tuned via Bayesian optimization. The preprocessing pipeline includes per-spectrum standardization, RobustScaler normalization of the target variables – effective temperature $T_{\mathrm{eff}}$, metallicity $[\mathrm{Fe/H}]$, and surface gravity $\log g$ – and data augmentation via Gaussian noise injection. On a held-out test set, the model achieved Mean Absolute Errors (MAE) of $59.76~\mathrm{K}$ for $T_{\mathrm{eff}}$, $0.103~\mathrm{dex}$ for $[\mathrm{Fe/H}]$, and $0.130~\mathrm{dex}$ for $\log g$. Normalized against the full-scale range of each parameter, these results represent range-normalized errors between $1\%$ and $3\%$, achieved with a highly efficient model complexity of approximately 540,000 trainable parameters. These results demonstrate that a compact residual multitask architecture, combined with principled signal preprocessing, provides a parameter-efficient solution for nonlinear parameter estimation in large-scale spectral datasets. In particular, the proposed model achieves competitive performance with substantially lower complexity than deeper neural network baselines.
arXiv:2606.11485v1 Announce Type: new Abstract: Mechanical strain is a powerful resource for tuning quantum systems, but existing piezoelectric strain cells are generally optimized for fragile, high-aspect-ratio single crystals rather than the thick, square-profile chips typical of semiconductor quantum devices. Furthermore, adapting these cells for qubits requires accommodating dense RF and DC wiring while maintaining strict electrical isolation from high-voltage piezo actuators. Here, we present a piezoelectric uniaxial strain cell designed to homogeneously strain thick, square-profile substrates. We introduce a highly symmetric dual-chip loading configuration that effectively suppresses flexural deformation and shear stress. The cell integrates a high-density RF/DC interposer to support standard wire bonding and encloses the actuators in a grounded Faraday cage to prevent unwanted Stark shifts in the device layer. Finite element simulations confirm that combining stiff actuators with this symmetric mounting drastically improves strain homogeneity. Finally, we validate the apparatus experimentally by applying uniaxial strain to a 200 $\mu$m thick silicon die. Surface strain measurements demonstrate an applied strain of 215 $\mu\epsilon$ for 200 V applied piezo bias.
arXiv:2606.11268v1 Announce Type: new Abstract: Understanding and forecasting lake dynamics is critical for monitoring water quality and ecosystem health across lakes and reservoirs. While machine learning methods have been recently applied to ecological time-series data, existing works assume regular sampling in time and depth, and struggle to generalize across lakes with heterogeneous variables, depths, and observation patterns. To address these limitations, we introduce \textsc{LakeFM}, a foundation model for aquatic systems, pre-trained on large-scale ecological datasets comprising both simulated and observed lakes. Through extensive empirical evaluation, we show that \textsc{LakeFM} learns meaningful representations spanning broader lake-level characteristics, and achieves competitive or often superior-forecasting performance compared to existing time-series foundation and non-foundation models, while producing physically plausible predictions consistent with real-world lake dynamics.
This study examines the effects of rural out-migration and remittance inflows on food consumption outcomes among rural households in the Tigray region of Ethiopia. Utilizing household survey data collected from 521 rural households across three distinct Weredas (districts) (Tahtay Maichew, Kola Tembien, and Kilte-awlaelo). A Binary Probit model was employed to identify factors influencing migration decisions, while an Endogenous Switching Regression (ESR) model was used to estimate the impact of migration on food consumption outcomes while controlling for selection bias and unobserved heterogeneity. Food security was measured using the Food Consumption Score (FCS) and dietary diversity indicators. The empirical results reveal that severe food insecurity is widespread, with over 60% of all surveyed households falling into the "Poor" food consumption category. Descriptive baseline comparisons show that migration and remittance transfers marginally shift the raw average FCS upward from 23.86 to 25.48. However, this impact is profoundly nuanced: remittances serve as an immediate consumption-smoothing safety net but run parallel to a "labor-lost" constraint that reduces own-production capacities, forcing households to rely increasingly on market purchases for staple foods. The findings reveal that migration creates short-term labor shortages in agricultural production; however, remittance inflows substantially improve household food consumption frequencies, particularly for pulses, vegetables, and other nutrient-rich foods. After accounting for self-selection bias and unobserved traits, the rigorous ESR estimates indicate that migration increases the Food Consumption Score of participating households by an average Treatment Effect on the Treated (ATT) of 10.75 points, shifting them into more secure dietary tiers. Moreover, remittances help households mitigate the adverse effects of drought and other shocks by relaxing liquidity constraints and supporting both food purchases and agricultural investments. The study recommends establishing target food security safety nets for non-remittance households, promoting scale-appropriate labor-saving agricultural technologies, expanding traditional communal labor-sharing innovations, and boosting irrigation and agricultural input support programs to enhance rural food security and livelihood resilience.
arXiv:2606.15271v1 Announce Type: cross Abstract: This work presents a transparent and reproducible benchmark study of a direct dual-network Physics-Informed Neural Network (PINN) formulation for the optimal control of a mass-spring-damper system. The classical linear-quadratic optimal control problem is solved by two independent classical methods – Pontryagin's Minimum Principle with single shooting, and direct transcription through trapezoidal collocation – and recast as a constrained optimization problem solved by two feedforward neural networks: a state network whose boundary conditions are enforced exactly through a composite cubic-and-mask ansatz, and an unconstrained control network. The composite loss combines the physics residual at the collocation points with a trapezoidal approximation of the cost functional, weighted by a single scalar hyperparameter. On the benchmark considered, the PINN reproduces the classical optimal cost to four significant digits, satisfies the terminal state constraints exactly by construction, and produces pointwise state and control errors that fall within the spread of the two classical references. Training is approximately two orders of magnitude slower than classical shooting on this benchmark, which is honestly reported. The contribution is methodological clarity rather than methodological novelty: the formulation and the accompanying Google Colab implementation are intended to lower the barrier to entry for practitioners exploring PINN-based optimal control without prior exposure to adjoint methods or two-point boundary value problems.
arXiv:2606.19535v1 Announce Type: cross Abstract: Large language models (LLMs) are increasingly deployed in sensitive settings such as software engineering, where their outputs directly shape downstream artifacts. Recent work has shown that an identical model can produce measurably different outputs depending on the deployment platform, a consequence of non-associative floating-point arithmetic and divergent kernel implementations. We study the security implications of this platform-dependent variability and uncover a novel attack surface on LLM deployments. We introduce FloatDoor, the first input-independent, platform-triggered backdoor attack against generative LLMs. The compromised model exhibits adversary-chosen behavior when served on a target platform and is otherwise benign. FloatDoor is realized through two lightweight LoRA adapters, one that amplifies inter-platform numerical divergence and one that binds the resulting platform signature to a malicious downstream task, while leaving aggregate model utility largely intact. FloatDoor exploits a pronounced time-of-check, time-of-use gap between model auditing and serving. We demonstrate FloatDoor on Qwen3-4B across a broad range of deployment targets, including NVIDIA GPUs, Google TPUs, AWS Graviton, and Alibaba Yitian-710. As a final case study, we show that FloatDoor reliably induces exploitable code vulnerabilities on a chosen target platform. Our results establish a new class of attacks on LLM deployments and underscore the pressing need for trusted model supply chains in sensitive, LLM-powered applications.
The deployment of LLM-based agents in scientific analysis raises opposing concerns: that agents may reduce methodological diversity, or that they may amplify the analytic flexibility through which researchers reach motivated conclusions. We argue these worries target two empirically separable layers: a design layer of methodological choices, and a verdict layer in which a decision rule maps estimates to a substantive claim. We test both by running 20 independent executions of Claude Code and Codex on a prominent immigration and social-policy against a many-analysts human baseline. At the design layer, Codex matches human methodological diversity and Claude Code produces nearly three times as many specifications; both agents' effect estimates remain broadly aligned with the human consensus, and no agent model exactly matches any human model. A prompt-induced anti-immigration researcher prior reorganizes each agent's methodological decisions but, unlike for biased human analysts in the same data, does not shift aggregate estimates or final verdicts; nor do agents reroute along the methodological axes humans use to bias their estimates. At the verdict layer, an explicit confirmatory prompt flips Claude Code's verdicts from 10% to 90% support while leaving its coefficient distribution essentially unchanged, operating through rule omission rather than rule softening. AI agents can rival or exceed human methodological diversity at the design layer while remaining vulnerable at the verdict layer. In our setting, the locus of AI bias is not estimation but interpretation.
Target Safety Assessment (TSA) requires systematic integration of genetic, transcriptomic, target homology, pharmacological, and clinical data to evaluate potential safety liabilities of therapeutic targets. This process is labor-intensive and expert-dependent, posing challenges in scalability and reproducibility. We present TSAssistant, a human-in-the-loop multi-agent framework that decomposes TSA report generation into a workflow of specialized subagents: Research Subagents that each ground and cite a single TSA domain, and Synthesis Subagents that integrate findings across domains. Subagents retrieve and synthesize evidence from curated biomedical sources through standardized tool interfaces and produce individually citable, evidence-grounded sections, with behavior shaped by a hierarchical instruction architecture that separates coordination logic from domain expertise and user intent. To complement these soft constraints, programmatic execution hooks and persistent memory stores enforce hard constraints across the workflow, while an interactive refinement loop allows experts to review and revise individual sections with full conversational context preserved across iterations. Rather than a single holistic comparison, we decompose report quality into reproducibility, evidential grounding, task-level accuracy, and controllability under expert oversight, finding high reproducibility and grounding, substantial agreement with the human reference, and net-positive expert-driven refinement.
arXiv:2606.13959v1 Announce Type: new Abstract: Sierra Leone's agriculture operates with almost no data-driven decision support, and no published machine learning study has examined the country's crop yields. We ask whether rice yield can be forecast from data Sierra Leone currently has. Using 25 years of FAOSTAT production data (2000-2024) for nine major crops, we train XGBoost, Gradient Boosting, and Random Forest under a strict anti-leakage protocol with expanding-window walk-forward evaluation across seven held-out years, benchmarked against naive persistence. No model trained on crop statistics alone outperforms persistence. Augmenting with free satellite climate data (CHIRPS rainfall, NASA POWER temperature) reverses this result: a climate-only XGBoost reduces forecast error by one third (RMSE 284 vs 428 kg/ha), a gain that holds for a linear model and is robust to excluding the anomalous 2018 season. Early-season (May-June) rainfall is the dominant predictor, implying seasonal yield risk is observable months before harvest. No model anticipated the 2018 collapse, whose origins were institutional rather than climatic. We translate the findings into policy recommendations for Sierra Leone's Feed Salone Strategy, with a fully open-source pipeline.
arXiv:2507.02921v4 Announce Type: replace-cross Abstract: Learning effective representations of urban environments requires capturing spatial structure beyond fixed administrative boundaries. Existing geospatial representation learning approaches typically aggregate Points of Interest (POIs) into pre-defined administrative regions such as census units or ZIP code areas, assigning a single embedding to each region. However, POIs often form semantically meaningful groups that extend across, within, or beyond these boundaries, defining places that better reflect human activity and urban function. To address this limitation, we propose PlaceRep, a geospatial representation learning method that constructs place-level representations by clustering spatially and semantically related POIs. PlaceRep summarizes large-scale POI graphs from U.S. Foursquare data to produce general-purpose urban region embeddings while automatically identifying places across multiple spatial scales. By eliminating model pre-training, PlaceRep provides a scalable and efficient solution for multi-granular geospatial analysis. Experiments using the tasks of population density estimation and housing price prediction as downstream tasks show that PlaceRep outperforms most state-of-the-art graph-based geospatial representation learning methods and achieves up to a x100 speedup in generating region-level representations on large-scale POI graphs. The implementation of PlaceRep is available at https://github.com/mohammadhashemii/PlaceRep.
Training psychotherapists in evidence-based interventions such as Acceptance and Commitment Therapy (ACT) requires repeated practice with meaningful feedback, yet opportunities for safe, standardized training are limited by ethical, logistical, and resource constraints. We introduce a system designed to support ACT-oriented psychotherapy training through spoken dialogue with an embodied virtual patient. The system uses large language models to simulate patient behavior conditioned on profiles derived from real therapy sessions and configurable clinical scenarios, while a separate automated evaluator provides turn-by-turn feedback on therapist responses based on established ACT fidelity criteria. Rather than aiming to replace supervision, the system is intended to support deliberate practice by enabling experimentation, reflection, and immediate feedback in low-risk settings. Expert evaluation with practicing psychologists confirmed high realism in patient behavior and demonstrated that immediate turn-by-turn ACT feedback increased therapists' awareness of intervention choices and enabled effective experimentation with alternative responses. Quantitative evaluation across 49 therapy transcripts identified GPT-4o-mini as the optimal feedback model, achieving the lowest mean absolute error (MAE = 6.12) in replicating human supervisor ACT fidelity ratings with statistically significant agreement. This work demonstrates the potential of fidelity-aware simulated patients as a scalable complement to psychotherapy training.
Object navigation in unseen indoor environments requires agents to perform semantic search under partial observability. Vision-language models (VLMs) provide strong semantic-spatial priors for this task, but how to interface them with robot navigation remains challenging: dense VLM inference is expensive, while abstracting environments into symbolic memories often separates high-level reasoning from the raw visual evidence that supports it. We propose we propose PIGEON (Point of Interest Guided Exploration for Object Navigation), a VLM-driven framework that formulates object navigation as raw-observation-grounded sparse decision problem. PIGEON introduces Points of Interest (PoIs) as sparse visual decision units that couple geometrically executable waypoints with raw egocentric observations. Rather than using VLMs as dense controllers or restricting them to frontier ranking, PIGEON enables VLMs to select among task-critical PoIs, including exploration frontiers, suspected target objects, traversable stairs, and floor-level summaries, while low-level planners execute continuous motion between them. This PoI interface further makes high-level navigation decisions verifiable, allowing us to develop an RLVR pipeline that improves local VLMs without manual Chain-of-Thought annotations. Extensive experiments on Habitat ObjectNav benchmarks show that PIGEON achieves state-of-the-art zero-shot performance, scales consistently with foundation model capacity, and transfers to Active Embodied Question Answering with only prompt modifications. Real-world deployments on physical robots further demonstrate its robustness and efficiency.
arXiv:2606.20227v1 Announce Type: new Abstract: Large Language Models (LLMs) have made significant progress in reasoning, particularly in deductive reasoning, which is crucial for high-stakes decision-making. As models improve, evaluation benchmarks should evolve to keep pace. However, existing benchmarks lack fine-grained control over logical complexity and struggle to balance semantic diversity with logical consistency. To address these issues, we propose QMFOL, an automated framework for generating monadic first-order logic reasoning tasks with quantifiable and controllable complexity. It constructs formal logical structures using conjunction and disjunction patterns, enabling precise control over reasoning depth, width, label types, and distractors. These structures are then translated into natural language via LLMs, with logical consistency ensured through round-trip verification using an external prover. Based on our framework, we build QMFOLBench, a benchmark comprising 2880 instances with 960 configurations across diverse logical and semantic dimensions. Evaluations on six large reasoning models (LRMs) and two LLMs show that performance degrades and computational overhead increases with rising logical complexity. Models perform better on True-labeled tasks than on False or Unknown ones, and exhibit sensitivity to semantic variation. Overall, QMFOL offers a scalable and reliable approach for constructing deductive reasoning benchmarks with controllable complexity, enabling more precise evaluation of reasoning capabilities in modern language models.
Possible explanations for the trend are emerging, but are likely to vary from one type of tumour to another. Possible explanations for the trend are emerging, but are likely to vary from one type of tumour to another.
arXiv:2606.13737v1 Announce Type: cross Abstract: Stream guardrails enable token-level safety detection before full responses are generated. However, they often make overly conservative judgements and block those sensitive but safe tokens, which is known as over-refusal. Due to lack of full context, they also fail to detect implicitly harmful content from jailbreaking. To address these challenges, we propose FreoStream, a novel streaming guardrail framework. Specifically, FreoStream fine-tunes a LoRA module to perform Future-Aware Reasoning when the base guardrail detects unsafe tokens. The reasoning process follows a Future-Reason-Judge paradigm: predict the future, reason about the full context and give the final judgement. This design can effectively reduce over-refusal by incorporating the future information. Moreover, we introduce the Safety-Aligned Optimization module that extracts the safety-aligned component from the reasoning gradients to update the base guardrail model, thereby enhancing streaming safety detection. Extensive experiments on various safety benchmarks demonstrate that FreoStream achieves lower over-refusal rates and better jailbreak defense compared to existing streaming guardrails.
arXiv:2606.17647v1 Announce Type: new Abstract: Quantum algorithms for integer factorization represent one of the most prominent applications of quantum computation, with far-reaching implications for modern cryptography. While Shor's algorithm provides a polynomial-time solution in the ideal quantum model, its practical implementation is severely constrained by the limitations of current noisy intermediate-scale quantum (NISQ) hardware. These constraints have motivated the exploration of alternative factoring algorithms with different structural and resource trade-offs. In this work, we present an experimental study of Regev's quantum factoring algorithm, implemented on real quantum hardware, and compare its behavior with that of Shor's algorithm under analogous conditions. Focusing on the case N = 15, we execute both algorithms on the QMIO quantum computer at the Centro de Supercomputacion de Galicia (CESGA) and contrast the results with one of IBM's open-access quantum computers and ideal simulations. This parallel execution enables a low-level comparison of the two algorithms, highlighting how their respective quantum implementations interact with hardware noise, limited circuit depth, and finite sampling. Our analysis emphasizes the different ways in which Shor's and Regev's algorithms encode arithmetic structure into quantum states through Fourier sampling in one and higher dimensions, respectively, and how these differences manifest in experimental outcomes. Although neither algorithm demonstrates a practical advantage in the small N regime, the results provide insight into their relative robustness and failure modes on contemporary quantum devices. This study illustrates the value of experimental benchmarking of alternative quantum factoring algorithms as a means of understanding the practical implications of algorithmic design choices in the NISQ era.
arXiv:2604.26963v2 Announce Type: replace-cross Abstract: Large language models (LLMs) are increasingly deployed as the execution core of autonomous agents rather than as standalone text generators. Agentic workloads induce a temporal shift from single-turn inference to multi-turn LLM-tool loops, and a spatial shift from chat-scale, GPU-only execution to repository-scale, GPU-CPU co-located execution. Consequently, coordinating heterogeneous resource demands of agentic execution has emerged as a critical system challenge. We design and implement MARS, an efficient and adaptive co-scheduling system that globally coordinates heterogeneous agentic workloads under coupled GPU-CPU resource pressure. By establishing holistic visibility across GPU inference and CPU tool execution via a unified information stream, an external control plane in MARS decouples admission from execution to prevent heterogeneous resource oversubscription. An internal agent-centric scheduler further minimizes the end-to-end critical path by prioritizing latency-sensitive continuations and adaptively retaining KV cache state only when warm resumption yields a latency benefit. Our evaluations show that MARS reduces end-to-end latency by up to 5.94x while maintaining nearly maximal system throughput. We further integrate MARS as the serving backend for the OpenHands coding agent framework, demonstrating its real-world effectiveness by accelerating end-to-end task completion time by up to 1.87x. Our source code is publicly available at https://github.com/Afterglow231/MARS_preview .
Human mitochondrial ATP synthase is an essential rotary motor enzyme that produces most of the cellular ATP through oxidative phosphorylation. Its membrane-embedded Fo sector contains highly hydrophobic transmembrane subunits that are challenging to study in aqueous environments without detergents. This study explores whether applying the QTY code can reduce the hydrophobicity of selected ATP synthase Fo subunits while preserving their overall molecular structures. We applied the QTY code to eight human ATP synthase Fo subunits: ATP6, ATP8, ATPK, ATP68, ATPMK, AT5G1, AT5G2, and AT5G3. Hydrophobic amino acids leucine (L), isoleucine (I), valine (V), and phenylalanine (F) in transmembrane regions were systematically replaced with hydrophilic glutamine (Q), threonine (T), and tyrosine (Y). Four native subunits with available CryoEM structures from human ATP synthase (PDB: 8H9S) were superposed with their AlphaFold3-predicted QTY analogs. The native ATP synthase Fo subunits superposed well with their respective QTY analogs. For the CryoEM-native comparisons, RMSD values ranged from 0.565[A] to 2.546[A]. For the AlphaFold3-native comparisons of subunits without CryoEM structures, RMSD values ranged from 0.204[A] to 0.297[A]. Despite substantial QTY substitutions in the transmembrane regions, ranging from 38.89% to 50.79%, the QTY analogs retained similar overall folds, molecular weights, and isoelectric points. Hydrophobic surface analysis showed that the QTY analogs had reduced hydrophobic patches compared with their native counterparts, with average hydrophobicity decreasing from 0.2959 in native proteins to -1.1023 in QTY analogs. These structural bioinformatics studies suggest that the QTY code can be applied to ATP synthase Fo subunits to generate more hydrophilic, potentially water-soluble analogs while preserving overall structural similarity. These results extend the application of the QTY code to the membrane-embedded Fo sector of ATP synthase and provide a foundation for future experimental studies testing whether these QTY analogs can be expressed, purified, and evaluated for assembly or proton-transfer-related functions.
arXiv:2606.17729v1 Announce Type: new Abstract: We prove almost tensorization for hypercontractivity and logarithmic-Sobolev constants for a class of reversible quantum Markov semigroups satisfying the positive off-diagonal scaling (PODS) property. This class includes qubit examples and generalized depolarizing semigroups with respect to full-rank states in arbitrary finite dimensions. For any such semigroup $(\Phi_t)_{t\ge 0}$ and every tensor power $n$, we show that the log-Sobolev constant of the product semigroup $\Phi_t^{\otimes n}$ is at least $2/(3\ln 2)$, approximately 0.96, times the log-Sobolev constant of the single-site semigroup $\Phi_t$, independently of $n$ and the local dimension $d$. The proof first establishes exact tensorization of the $(q,2)$-hypercontractive inequality for integer $q$, in particular $q=3$, and then extends the estimate to all real $q>2$ by complex interpolation; the standard implication from hypercontractivity to logarithmic-Sobolev inequalities yields the stated almost tensorization result. As an application of the same method, we also obtain sharp $(q,2)$-hypercontractivity estimates for qubit depolarizing channels.
When vision contradicts text, multimodal large language models (MLLMs) consistently favor text, even when images provide clear evidence otherwise. This bias poses risks for applications requiring visual grounding, yet its cause remains unclear. In this paper, we uncover a surprising finding: models often get it right initially, forming correct vision-based predictions in their intermediate layers, before changing their minds and favoring text in the final output. We call this "late-layer textual override". The visual information is encoded, it simply does not survive to the output. More intriguingly, we find that how predictions change reveals whether they're correct: 85% of failures shift toward text, while 89% of successes shift toward vision. This directional signature enables a simple but powerful intervention: when we detect a confident visual prediction being suppressed, we restore it. We propose CALRD (Conflict-Aware Layer Reference Decoding), a training-free method that recovers overridden predictions at inference time. Experiments across five MLLMs of varying architectures demonstrate up to 9.4% absolute improvements on conflict benchmarks while largely preserving standard performance, without training or external knowledge. It recovers what the model already knew but failed to preserve.
arXiv:2606.20554v1 Announce Type: cross Abstract: Generative recommendation is an emerging paradigm that has shown promise in industrial recommendation systems, aiming to predict users' next interactions from their historical behaviors. At the core of generative recommendation lies item tokenization, which bridges item semantics and recommendation models. However, existing methods often struggle to effectively organize and inject complex user-behavioral and item-semantic contexts into recommendation models simultaneously. On the one hand, existing graph-based integration methods, such as graph serialization and graph neural networks, either suffer from scalability issues or exploit only local graph information. On the other hand, existing semantic tokenization methods typically rely on heuristics and lack explicit supervision signals, which may lead to inaccurate or suboptimal semantic representations. To address these limitations in user interest context modeling, we propose G2Rec, a scalable framework that unifies holistic graph-based user co-engagement modeling with semantic tokenization for industrial-scale generative recommendation. Overall, G2Rec enables recommendation models to capture holistic and semantically grounded user interest prototypes without requiring ground-truth user interests, thereby providing more comprehensive and accurate modeling of user behavior contexts in industrial sequential recommendation. Online deployment across product surfaces and extensive experiments on public datasets demonstrate the superiority of G2Rec over existing methods.
AI coding agents have rapidly transformed software engineering, powering widely used interactive coding assistants. Despite their interactive real-world use, existing benchmarks evaluate them as fully-autonomous systems. In this work, we introduce Dialogue SWE-Bench, an automatic benchmark dataset for evaluating the ability of coding agents to resolve real-world software engineering problems through dialogue with a user. We design a novel, persona-grounded user simulator to support our task evaluation, and augment our task evaluation with automatic evaluations of dialogue quality. We also propose a new schema-guided agent, aimed at improving the dialogue capabilities of off-the-shelf coding agents, which improves over strong baselines by 3-14%. Our results indicate that better coding models do not always correspond to better dialogue models, suggesting that dialogue capability is a distinct and currently understudied dimension of coding agent performance.
Structured width pruning of GLU-MLP layers in Llama-3.2 models, guided by the Peak-to-Peak Magnitude (PPM) criterion, reveals a systematic dichotomy in how reducing the expansion ratio affects different model capabilities. While performance on tasks relying on parametric knowledge (e.g., MMLU, GSM8K) and perplexity metrics degrades predictably with decreasing expansion ratios, instruction-following capabilities improve at the 2.4x equilibrium ratio (IFEval: +4.8 points / +46% in Llama-3.2-1B and +3.7 points / +39% in Llama-3.2-3B), and multi-step reasoning remains robust (MUSR). This pattern, observed consistently across both evaluated model sizes, challenges the prevailing assumption in compression research that pruning induces uniform degradation. To investigate this, we evaluated seven expansion ratio configurations using comprehensive benchmark suites that assess factual knowledge, mathematical reasoning, language comprehension, instruction-following, and truthfulness. Our analysis identifies the expansion ratio as a critical architectural parameter that selectively reshapes the model's task performance profile, rather than merely serving as a compression metric.
End-to-end task-success is the dominant way to evaluate LLM agents, but one aggregate number tells you that an agent regressed, not where. We present layer-isolated evaluation: a deployed ordering agent is decomposed into a fixed taxonomy of layers (ontology, intent, routing, decomposition, escalation, safety, memory, and cross-cutting envelope/defense), each exercised by its own assertion slice in a deterministic, no-LLM "pure" mode. The pure suite (238 cases across 23 slices; 225 run in 2.39 s, ~10 ms/case) runs in CI on every change against a locked per-slice baseline. We validate by controlled regression injection, degrading one layer at a time across seven non-safety layers. The effect we did not design in is masking: the aggregate pass-rate barely moves (-1.7 to -5.9 pp for six local regressions), while the matching slice craters (-25 to -91 pp). A layer's slice reacting to its own fault is partly by construction; the measured results are (i) the aggregate masking and (ii) that damage stays off the other slices: the injected layer's slice is the single worst-hit in 5 of 7 cases and top-3 in 7 of 7 (mean rank 1.29 of 19). Localization replicates on a second, structurally different tenant (Starbucks SG): all seven matching slices crater, so it is not a single-catalog artifact. We position it as a concrete, deterministic instantiation of the component-level evaluation EDDOps prescribes but leaves unimplemented, with CheckList as ancestor and as the deterministic mirror image of whole-workflow stochastic mutation testing. Our contributions: (a) a fully decomposed, sub-second, no-LLM per-layer harness for a production agent, (b) a coverage-honesty test-adequacy criterion that refuses to score an unexercised layer, and (c) the regression-injection demonstration that per-slice baseline-locked gates localize regressions an aggregate metric masks.