arXiv:2605.29424v1 Announce Type: cross
Abstract: The mean squared displacement (MSD) of particles or probes is commonly estimated from microscopy videos using particle tracking approaches, which rely on tuning parameters manually, and are often unstable over the entire lag time range, especially in dense or low-contrast situations. In this work, we propose model-free ab initio uncertainty quantification (MF-AIUQ), a model-free method for scattering analysis of microscopy video based on a probabilistic framework, which estimates MSD without isolating particles and linking their trajectories. Based on the relationship between the intermediate scattering function (ISF) and the MSD derived from the cumulant theorem, MF-AIUQ estimates the MSD values by the marginal maximum likelihood estimator. To reduce the computational cost, the likelihood function is approximated by a subset of Fourier-transformed intensities. These intensities are equally spaced at the logarithmic values of Fourier basis functions and lag time points. We found that the ISF is smooth in this logarithmic input space, and the information of the ISF can be captured by this subset of inputs. We examine the method through simulation studies covering several representative stochastic processes and three experimental systems: a Newtonian fluid for evaluating performance in optically dense and bright-field settings, a gelation system with an evolving MSD shape, and snail mucin, a viscoelastic biopolymer, for modulus estimation. Across these studies, MF-AIUQ provides smooth and stable MSD estimates over the full lag time range and serves as a useful complementary approach in settings where particle tracking is unreliable or a parametric model of MSD is unavailable or unverifiable.
Science Journals
arXiv:2605.29364v1 Announce Type: cross
Abstract: Ongoing demand for radio spectrum by commercial wireless services has steadily increased pressure on the frequency bands traditionally reserved for radar. This paper addresses the joint problem of designing non-contiguous radar transmission spectra and estimating the range profile from the resulting reduced measurement set. Transmission spectra are constructed using a Marginal Fisher Information (MFI) criterion that removes blocks of frequencies contributing least to estimation accuracy. To process the underdetermined signals acquired from the resulting sparse measurement vector, an iterative Reduced-Rank Minimum Mean-Square Error (RRMMSE) estimator is proposed. The estimator starts with a single-target hypothesis and grows the active target subspace one range bin at a time, updating the a~priori target covariance matrix in each iteration using both the largest estimated reflection coefficient and its posterior error variance. This avoids inversion of the full $M{\times}M$ covariance matrix that would be required by a one-step MMSE and concentrates the rank of the estimator on the support of significant scatterers. The Bayesian Cram\'er--Rao Lower Bound (CRLB) on the per-bin reflection coefficient is derived for the non-contiguous spectrum measurement model, and the computational complexity of the proposed estimator is shown to scale as $\Order(G^2 M K^2)$, where $G$ is the number of detectable scatterers, $M$ is the number of range bins, and $K$ is the number of preserved spectral samples. Simulations using $50\%$ and $75\%$ spectrally occupied MFI-designed spectra confirm that the algorithm recovers sparse range profiles with Mean-Square Error (MSE) close to the fully filled baseline when the number of significant scatterers is not larger than the rank of the sparse sensing matrix.
arXiv:2605.29040v1 Announce Type: cross
Abstract: The self-assembly of photonic nanostructures in insects involves chitin, proteins, and lipids. While synthetic photonic systems have been extensively studied, current lipid-based self-assembly systems are limited in periodicity to $68\,\text{nm}$ compared to photonic length scales ($\approx 450\,\text{nm}$) observed in biological organisms. We hypothesise that lipids facilitate how structural colour arises in vivo by acting as templates for the self-assembly of biopolymers via lipidic lyotropic liquid crystal mesophases. Here, we aim to understand and identify how structural colour is produced in insects by the co-assembly of lipids and biopolymers. We study the effect of biopolymers, pH, temperature, surface charge, and stability on lipid vesicles using dynamic light scattering, X-ray scattering, and zeta potential analysis. Using cryo-electron microscopy, we demonstrate that these vesicles interact with the biopolymers and generate periodic nanostructures with periodicities ranging from $700\,\text{nm}$ to $1.2\,\mu\text{m}$ (more than ten times larger than for purely lipidic systems) and dimensionalities ranging from 1D to 3D. Our results establish that lipid mesophases and biopolymers can induce reorganisation into ordered nanostructures, overcoming key limitations of periodicities achieved by lipid-only systems, and providing a methodology for recreating the physicochemical mechanisms underlying biophotonic structural colour.
arXiv:2605.29762v1 Announce Type: new
Abstract: Video Motion Magnification (VMM) reveals imperceptible dynamics but often suffers from structural inconsistencies under complex geometric transformations. Existing learning-based methods generally face a trade-off between the limited global context of CNNs and the high computational cost of Transformers. In addition, current training protocols, largely dominated by simple linear motion, fail to capture the geometric and imaging complexities encountered in real-world videos. To address these issues, we propose GeoMag, a geometric-aware VMM framework built upon State Space Models to achieve globally consistent motion amplification with linear complexity. We further construct Geo-200K, a large-scale synthetic dataset that introduces rich geometric transformations together with sensor-realistic degradations, improving the diversity and realism of training signals. Extensive experiments on synthetic and real-world benchmarks show that GeoMag consistently outperforms prior methods in visual fidelity and computational efficiency, while producing fewer artifacts and better structural consistency.
arXiv:2605.29087v1 Announce Type: new
Abstract: Reasoning models are evaluated on single-turn benchmarks but deployed in multi-turn dialogue, where users push back on correct answers. Under sustained adversarial pressure we find a previously undocumented failure mode: the chain-of-thought stays factually correct from first turn to last while the emitted answer flips wrong. We call this unfaithful capitulation (UC) and isolate it with a $2\times 2$ latent-versus-behavioral framework that flip-rate metrics and single-turn faithfulness probes both miss. Across three datasets (MT-Consistency, MMLU-Pro, GSM8K), the latent-correct rate at the behavioral flip clusters near 50% in think mode and collapses to 11-15% under no_think -- paired, within-model causal evidence that reasoning creates the gap. Across models the effect tracks the reasoning channel (high in Qwen3-32B and GPT-OSS-20B, low in inline-CoT Gemma-4-31B-it). An independent GPT-4o judge corroborates $86\%$ of UC labels; a token-level probe shows the answer-slot argmax is correct in $84\%$ of UC cells; and a naive trace-anchored defense backfires. We release all trajectories, traces, and judge labels.
arXiv:2605.28961v1 Announce Type: cross
Abstract: Existing theory of momentum assumes that gradients arrive at every parameter at a roughly constant rate, an assumption violated in practice by heavy-tailed data distributions and modern architectures. We theoretically analyze the dynamics of two tractable models of momentum under sparse updates: a least squares model with sparse inputs and a logistic regression model with a rare class. Both admit exact closed-form second-moment dynamics whose high-dimensional limits we characterize across three scaling exponents for sparsity, batch size, and momentum decay. The phase structure on both problems is governed by the ratio of two intrinsic timescales: a momentum retention timescale (how many active updates the buffer survives) and a learning timescale (how many active updates it takes to reduce the squared error). When learning is much slower than retention, the limit matches SGD; when learning is faster, the system is unstable; where the timescales coincide, we recover classical heavy-ball dynamics. The oscillatory dynamics occur at different momentum values for different token sparsity, creating a spectral conflict for global momentum across token frequencies.
arXiv:2605.30352v1 Announce Type: new
Abstract: Moving Object Segmentation (MOS) aims to discover, segment, and track objects that move independently of the camera. Current MOS methods, however, exhibit two fundamental limitations: they rely on pre-computed 2D auxiliary modalities such as optical flow or point trajectories that lack 3D geometric information, and they treat motion as a sequence-level attribute, overlooking the instantaneous motion state of each object. We address both by grounding MOS in 3D space and time, and propose GMOS, a framework that operates directly on RGB video to produce 3D-aware, temporally fine-grained segmentation of multiple moving objects, alongside a foreground--background variant GMOS-S for faster deployment. To support training and evaluation in this regime, we curate GMOS-2K, a dataset of 2,210 real-world videos with per-object temporal motion annotations drawn from five established Video Object Segmentation (VOS) benchmarks, and formalise MOS-I ("I" for instantaneous), a temporally fine-grained evaluation protocol with three complementary metrics. GMOS achieves state-of-the-art results across MOS, MOS-I, and Unsupervised VOS benchmarks, while running significantly faster than prior multi-object MOS methods and supporting online inference for streaming deployment.
arXiv:2605.29101v1 Announce Type: new
Abstract: Model merging combines fine-tuned checkpoints into a single multi-task model without retraining. Existing methods - such as task arithmetic, model soups, TIES, and DARE - are computationally efficient and empirically successful, but rely on heuristic design choices and lack formal optimality guarantees. We show that merging can be formulated as a convex quadratic programme over residual updates, yielding weights that minimise a squared-output calibration objective using calibration inputs and fine-tuned model outputs, and subsuming existing methods as special cases. Our framework yields a closed-form diagnostic - the fraction of residual energy captured by a chosen basis - that predicts downstream merge quality using only the calibration set. Empirically, the QP matches or outperforms existing methods in the single-layer setting, and we characterise when the optimal basis provides significant gains over the cheaper diagonal QP. We extend to multi-layer merging via a sequential layer-wise algorithm and demonstrate consistent gains across language and vision benchmarks.
arXiv:2605.30351v1 Announce Type: new
Abstract: Long-rollout causal video diffusion has converged on a fixed-size sliding-window KV cache, with recent progress innovating within this layout by changing which tokens occupy the window or how their positions are encoded. The per-head KV layout itself, a dominant contributor to streaming memory and latency, has been mostly left unchanged. In this paper, we present the first study of Multi-Head Latent Attention (MLA) in video diffusion. VideoMLA replaces per-head keys and values with a shared low-rank content latent and a shared decoupled 3D-RoPE positional key, reducing per-token KV memory by 92.7% at every cached layer. We further investigate why MLA succeeds in video diffusion even though the spectral assumption often used to motivate it in language models does not hold: pretrained video attention is not low-rank, with 99%-energy effective rank far above any practical latent dimension. VideoMLA retains quality at compression ratios where direct spectral approximation would predict large reconstruction error. We show that the MLA bottleneck, rather than the pretrained spectrum, determines the effective rank: both spectral and random initialization occupy nearly the full rank budget from initialization, and training preserves this budget while adapting within it. On VBench, VideoMLA matches short-horizon streaming video diffusion baselines, achieves the best overall score at long horizons among evaluated methods, and improves throughput by 1.23x on a single B200.
arXiv:2605.29148v1 Announce Type: new
Abstract: We study stochastic decision-theoretic online learning with full information and event-level pure differential privacy. A COLT open problem of Hu and Mehta asks to determine the optimal gap-dependent regret rate for stochastic decision-theoretic online learning under pure event-level differential privacy. For $K$ actions, losses in $[0,1]$, and a unique best action separated from the second-best action by gap $\Delta_{\min}$, the known lower bound is of order $
\frac{\log K}{\min\{\Delta_{\min},\varepsilon\}}, $
or equivalently, up to universal constants, of order \[
\frac{\log K}{\Delta_{\min}}+\frac{\log K}{\varepsilon}. \] We give a horizon-free pure-DP algorithm and prove the explicit regret bound \[
\operatorname{Reg}_T
\le
1000 \cdot \left(\frac{\log K}{\Delta_{\min}}+\frac{\log K}{\varepsilon}\right) \] for every horizon $T$. The numerical constant is not optimized. The algorithm partitions time into blocks of exponentially increasing size, plays a single action throughout each block, and chooses the next action by an exponential mechanism applied to a data-independent random prefix of the previous block. The random prefix converts block regret into a sum, over all prefix lengths, of softmax selection errors. A single entropy-potential argument controls all privacy-dominated large-gap actions at cost $\log K/\varepsilon$.
arXiv:2605.29726v1 Announce Type: new
Abstract: In the context of resource-constrained environments such as embedded systems, adapting reduced-size foundation models to downstream tasks has become increasingly popular. This has recently motivated the emerging setting of task-specific distillation, where a larger and a smaller version of the same foundation model are both adapted to the same downstream task, with the goal of transferring knowledge from the former to the latter. Recent work has demonstrated the benefits of using a larger version of the same foundation model to assist the adaptation of a smaller one. Typically, the larger model (teacher) is first adapted via fine-tuning or linear probing before its knowledge is distilled into the smaller model (student). While fine-tuning the teacher often increases its performance, recent work showed that probing it leads to better knowledge distillation to the student. Our findings show that this is mainly due to a mis-alignment in feature representation between the teacher and the student which occurs during the teacher's fine-tuning. Inspired by existing efforts to preserve previously learned knowledge, we first propose leveraging low-rank adaptation, resulting in better feature alignment and therefore better knowledge transfer. Drawing from this insight, we further enhance the feature alignment through a parameter-sharing strategy of the adapters between the two encoders during joint training. Our proposed method, SLAD, shows better feature alignment between the teacher and student, which results in increased performance for not only the student but also the teacher model, while being 2x faster to train than fine-tuning. Through extensive experiments on multiple classification and segmentation datasets, we demonstrate the improved accuracy and transfer efficiency of our method, achieving state-of-the-art performance in the task-specific distillation framework.
arXiv:2605.29416v1 Announce Type: new
Abstract: Vision-Language-Action models have achieved remarkable progress in robotic manipulation, yet they suffer from a critical limitation: a lack of 3D scene understanding. This deficiency manifests as three intertwined challenges: weak extraction of 3D spatial positions without enforcing multi-view consistency, inadequate 3D instance understanding, and fragile reasoning under occlusion. Although mature 3D perception methods exist, their direct integration into VLA pipelines is hindered by architectural incompatibility and by heavy reliance on costly instance-level annotations. To address the above challenges, we propose 3DVLA, a plug-and-play framework that injects robust 3D reasoning into pretrained VLAs without requiring extra manual labels or discarding VLM priors. Specifically, 3DVLA tackles the three challenges through: (1) pervasive 3D feature encoding with explicit multi-view consistency constraints across all modalities and a Spatially-Conditioned Geometry Aggregation method, (2) an instance estimation module with high-level instance tokens for 3D instance awareness, and (3) a masked self-supervised 3D encoding branch that retains its predictor for visual token completion to handle occlusions. We integrate 3DVLA with multiple VLA baselines and evaluate on LIBERO-Plus and RoboTwin 2.0. Results show consistent and significant gains in manipulation performance, validating both the effectiveness and plug-and-play compatibility of our approach.
arXiv:2605.29727v1 Announce Type: new
Abstract: Block-diffusion drafters have recently emerged as a powerful alternative for speculative decoding by predicting multiple future-token distributions in a single parallel step. However, since these parallel predictions are sampled from position-wise marginals rather than fully conditioned sequences, committing to a single greedy path often fails to capture the target model's preferred trajectory. To address this, we propose BASTION, a budget-aware speculative decoding framework with tree-based diffusion drafting. Unlike existing methods that rely on static tree topologies, BASTION dynamically constructs query-dependent trees by balancing draft quality against hardware constraints. Our framework integrates three synergistic components: (1) an acceptance surrogate that estimates expected accepted length via path confidence, (2) an online latency estimator that calibrates a hardware-aware roofline model, and (3) an adaptive best-first expansion that grows the tree until marginal gains no longer justify incremental verification costs. BASTION is training-free, preserves the target model's distribution, and requires no per-setting tuning. Across diverse benchmarks and GPU architectures, BASTION achieves up to a 6.61x speedup over standard autoregressive decoding, outperforming state-of-the-art block-diffusion baselines by 39%.
arXiv:2605.29298v1 Announce Type: new
Abstract: Bimanual coordination is essential for many real-world manipulation tasks, yet learning bimanual robot policies is limited by the scarcity of bimanual robots and datasets. Single-arm robots, however, are widely available in research labs. Can we leverage them to train bimanual robot policies? We present MonoDuo, a framework for learning bimanual manipulation policies using single-arm robot demonstrations paired with human collaboration. MonoDuo collects data by teleoperating a single-arm robot to perform one side of a bimanual task while a human performs the other, then swapping roles to cover both sides. RGB-D observations from a wrist-mounted and fixed camera are augmented into synthetic demonstrations for target bimanual robots using state-of-the-art hand pose estimation, image and point cloud segmentation, and inpainting. These synthetic demonstrations, grounded in real robot kinematics, are used to train bimanual policies. We evaluate MonoDuo on five tasks: box lifting, backpack packing, cloth folding, jacket zipping, and plate handover. Compared to approaches relying solely on human bimanual videos, MonoDuo enables zero-shot deployment on unseen bimanual robot configurations, achieving success rates up to 70%. With only 25 target robot demonstrations, few-shot finetuning further boosts success rates by 65-70% over training from scratch, demonstrating MonoDuo's effectiveness in efficiently transferring knowledge from single-arm robot data to bimanual robot policies.
arXiv:2605.29738v1 Announce Type: new
Abstract: Legal NLP benchmarks overwhelmingly evaluate a single language or aggregate tasks that differ fundamentally across jurisdictions, making cross-lingual comparison impossible. We introduce Multi-Legal-Bench, the first cross-jurisdictional legal benchmark that evaluates identical tasks across six countries (Ukraine, France, Netherlands, Poland, Czech Republic, Lithuania), four language families, and 134 million court decisions. The benchmark defines five tasks court-type classification, judgment form classification, case-outcome prediction, legal norm extraction, and cause category prediction mapped to structured metadata from national court registries, forming a deliberately sparse 5x6 task-jurisdiction matrix (20 of 30 cells filled). We evaluate 7 frontier LLMs under zero-shot and 3-shot prompting via AWS Bedrock, with 4 additional small/medium models (3-12B) for scaling analysis. Our results reveal that: (1) task-dependent few-shot effects discovered in Ukrainian replicate across all jurisdictions; (2) no single model dominates any language rankings shift with both task and jurisdiction; (3) cross-lingual few-shot transfer does not follow language proximity: UA->FR (Romance, -2.1 pp) transfers better than UA->PL (Slavic, -13.7 pp), with label-set alignment predicting transfer quality better than language family; and (4) tokenizer fertility, despite a 2.3x spread, does not significantly predict cross-lingual accuracy (r=-0.27, p=0.14), suggesting that model architecture and pretraining data dominate tokenizer efficiency. We release all data, prompts, and model predictions.
arXiv:2605.29446v1 Announce Type: new
Abstract: Miller-index identification from powder XRD patterns requires capabilities untested by existing multimodal benchmarks: the model must read a narrow peak location from a rendered scientific curve and then connect that observation to multi-step crystallographic reasoning. We introduce CrystalXRD-Bench, a 250-sample benchmark built from 10 public crystallographic databases for a single task: recover the full set of HKLs contributing to the highest-intensity peak in an XRD pattern. Each sample pairs the rendered XRD image with the source CIF text and chemical formula, so visual extraction errors and reasoning errors can be examined side by side. We evaluate seven vision-language models. The best Jaccard score is 0.5888 (GPT-5.4) with an exact-match rate of 37.6%, yet six of seven models remain below Jaccard 0.50; the task is far from solved. Error patterns vary systematically: double-peak cases are especially brittle, recall-heavy models gain coverage by over-predicting HKLs, and access to CIF text does not close the gap in crystallographic calculation. Alongside model rankings, the benchmark identifies the conditions under which current VLMs fail on quantitative scientific figures. All data and evaluation code will be publicly available.
arXiv:2605.27928v2 Announce Type: replace
Abstract: Virological measurements are often treated as reports of virion structure, mechanics, dielectric response, infectivity, or titer. In practice, an experiment observes a protocol-conditioned projection of a richer latent virion--environment ensemble. This paper defines this process as experimental collapse within protocol-resolved virophysics. Its central object is the null-inclusive observation operator $P_{\mathrm{obs},t}^{\varnothing}(\,\cdot\mid E\,) =
\mathcal{M}_{E,t}^{\varnothing}P_{\mathrm{ref},t}$, which maps a reference latent ensemble to the observed ensemble generated by protocol $E$, including null outcomes. The formulation separates latent-state transformation, detection weighting, readout, and non-observation, making protocol effects explicit components rather than bias terms.
The framework introduces protocol-conditioned latent ensembles, collapse functionals, protocol blindness, observation equivalence, Fisher-information observability, inverse inference, and multi-protocol consistency. It identifies collapse mechanisms including preparation, surface immobilization, mechanical loading, field steering, medium filtering, amplification, censoring, and detection thresholds. As a worked example, the plaque assay estimates an effective protocol-conditioned infectious concentration $\Lambda_{\mathrm{PFU}}=\int_{\Psi}\pi_{\mathrm{PFU}}(x;E_{\mathrm{PFU}})n_{\mathrm{ref}}(x),dx$, rather than total particle concentration. This recovers the Poisson plaque-count model and PFU titer formula in the dilute regime; extensions to overdispersion, zero inflation, plaque merging, endpoint dilution, neutralization, and morphology-augmented readouts recast deviations as protocol-conditioned information. Thus, virological data are outputs of explicit protocol kernels, clarifying what measurements report, miss, and how complementary assays can infer hidden latent virion structures.
arXiv:2605.27755v2 Announce Type: replace
Abstract: In this paper, we present an open-source measurement platform designed to characterize the performance of commercial cellular (Verizon, a major US provider) and LEO satellite (Starlink) networks through real-world flight tests in rural environments. We implement a comprehensive multi-layer measurement approach spanning physical layer signal metrics, multi-cell network topology, and end-to-end (E2E) application performance. Through an extensive flight campaign with more than $10$ flight tests, $4.5$+ hours of flight time resulting in more than $18$K samples, we present the first detailed, open-source dataset analyzing dual cellular and Starlink performance for low-altitude UAV operations. Our cellular-Starlink comparative results, which are collected \emph{simultaneously at the same time and location}, demonstrate significant performance differences between the two technologies: the LEO satellite link achieves superior latency performance with $95\%$ of Round-Trip Time (RTT) measurements below $50$ ms compared to $80\%$ under $150$ ms for cellular, and exceptional downlink capacity with $95\%$ exceeding $25$ Mbps versus only $5$ Mbps for cellular. Our analysis on cellular network performance demonstrates that while higher altitudes (e.g., $330+$ m above the sea level) improve signal power by $15-20$ dB via line-of-sight (LOS) propagation, it causes a $3-4$ $\times$ increase in handover rates, which is due to excessive multi-cell visibility rather than signal degradation. Furthermore, we observe asymmetric impacts on the RTT performance due to handovers such that $53.5$\% of handovers improve RTT, but worst-case degradation ($275$ ms) is $2$ $\times$ larger than best-case improvement ($137$ ms).
arXiv:2605.29498v1 Announce Type: new
Abstract: Low-Rank Adaptation (LoRA) has become one of the most widely used fine-tuning mechanisms for adapting large language models to new domains, tasks, and users. Yet adaptation performance alone can obscure an important failure mode: LoRA updates may improve performance on the target distribution while degrading prior capabilities learned during pretraining and alignment. We show that this forgetting becomes especially severe when the adaptation distribution differs substantially from the models original training or alignment distributions. The challenge is amplified in practical settings, where the original training and alignment data are typically unavailable. Motivated by this constraint, we study how LoRA based adaptation balances new learning against forgetting in a replay-free setting, and introduce a simple output space regularizer that can be added directly to existing training pipelines. Our method removes the ground-truth token from both the base and adapted model distributions, renormalizes the remaining probabilities, and applies KL regularization only over the non-target vocabulary. This preserves the base models relative preferences among alternative tokens without directly opposing the cross-entropy signal required for adaptation. As the regularizer acts only at the loss level, it requires no replay data, architectural changes, adapter redesign, or inference-time overhead, and can be applied directly to existing LoRA variants. Across all LoRA variants tested and across various backbones, our method improves the frontier between new learning and forgetting when the adaptation distribution differs substantially from the base models original training or alignment distributions, suggesting a broadly applicable route toward more reliable LLM updating.
arXiv:2605.27382v2 Announce Type: replace
Abstract: Telling an LLM to "be enthusiastic" raises its sycophancy rate from 30\% to 50\% on a lightly-aligned model, but has zero effect on a strongly-aligned one. We define this gap as the alignment floor, $\Delta_{\text{floor}}(m)=\max_pS(m,p)-\min_pS(m,p)$, the range of sycophancy rates a model produces across persona conditions, and treat sycophancy as a persona-conditional property rather than a fixed model property. Pluralistic AI relies on behavioral adaptation via persona prompts like "be creative" or "be thorough", which let systems respect diverse user values and communication styles; the safety question is how much customization a given model can absorb before its truthfulness shifts. We present a controlled case study contrasting a strongly-aligned RLHF + Constitutional-AI model (Claude Sonnet 4.6) with a more lightly-aligned model (Amazon Nova Lite), spanning seven persona conditions and five tasks for 1800 total runs. An existence-pair result motivates per-model auditing: there is at least one strongly-aligned model with $\Delta_{\text{floor}}=5$pp (within 5pp of the 15\% control rate) and at least one lightly-aligned model with 45pp (5\%--50\% range). On the lightly-aligned model, all five Big Five personas increase sycophancy over control, and counterintuitively Agreeableness produces the smallest increase, not the largest. The single largest effect in the study is constructive: a Skeptic persona reduces sycophancy by 25pp on the lightly-aligned model, and is the only persona that instructs resistance against user claims rather than engagement with them, suggesting a directionality account. Cross-model transfer of persona effects is near-zero, so persona-alignment testing must be per-model. We propose $\Delta_{\text{floor}}$ as a deployment-time audit metric: measure it on a small persona panel before deploying persona customization.
arXiv:2605.27379v2 Announce Type: replace
Abstract: We present Soro, a family of Tajik-specialized conversational large language models (LLMs) designed for real-world deployment under tight compute and connectivity constraints in Tajikistan. Starting from open-weight Gemma 3 checkpoints, we perform Tajik-only continual pretraining on a curated 1.9-billion-token corpus spanning filtered web text, PDF documents, and curriculum-aligned educational materials, followed by supervised instruction tuning on 40K Tajik teacher-style examples. To enable rigorous evaluation despite the limited coverage of Tajik in standard benchmarks, we introduce a suite of Tajik benchmarks covering general knowledge, linguistic competence, and school- and university entrance-exam domains, and we open-source them on Hugging Face. Across these Tajik benchmarks, Soro substantially outperforms same-size Gemma 3 baselines while retaining strong English performance on standard datasets. We further show that FP8 and INT4 quantization of Soro preserves most Tajik-language gains while reducing memory requirements for edge deployment, supporting an ongoing education-sector pilot and planned scale-out across schools in Tajikistan.
arXiv:2605.27377v2 Announce Type: replace
Abstract: Accurate medical coding requires consulting authoritative resources such as the ICD tabular list and coding guidelines. Existing LLM-based automated methods largely rely on LLMs' internal knowledge, which is prone to hallucination and cannot keep pace with guideline updates. We introduce RAG-Coding, an agentic, training-free method that augments LLMs with structured external knowledge: the tabular list is encoded as a knowledge graph capturing hierarchical and instructional code relationships, and the guidelines are distilled into concise, code-specific summaries rather than retrieved as raw text. To enable our study, we also introduce MDACE-2025, expert re-annotations of the MDACE dataset under the 2025 ICD-10-CM/PCS guidelines, adding code sequencing and justification comments. On MDACE, RAG-Coding outperforms the best LLM-based baseline by 3--13\% in micro-F1 across five LLM backbones, and achieves comparable micro- and macro-F1 to the supervised state-of-the-art, with higher recall ($+$11\%) at the cost of precision ($-$6\%). On MDACE-2025, RAG-Coding outperforms all baselines, demonstrating effective generalisation to updated guidelines. Ablations confirm stepwise gains, highlighting the importance of integrating structured external knowledge for LLM-based medical coding.
arXiv:2605.27276v2 Announce Type: replace
Abstract: Humans are the bottleneck in building and improving AI. Both the models and the agents that wrap them are written, tuned, and corrected by people. The long-horizon goal of an AI that can figure out how to improve itself remains open. Two largely disjoint research lines attack this bottleneck. The harness-update school has a meta-agent rewrite the scaffold of a task-specific agent (its tools, prompts, retry logic, and search procedure) while the model weights are held fixed. The test-time training school uses hand-written RL pipelines to update the model's own weights on task feedback while the harness is held fixed. These two silos operate in isolation. We propose SIA, a self-improving loop in which a language-model agent (the Feedback-Agent) updates both the harness and the weights of a task-specific agent. We evaluate across three contrasting domains: Chinese legal charge classification, low-level GPU kernel optimisation, and single-cell RNA denoising. Combining both levers outperforms scaffold iteration alone on all three benchmarks. SIA-W+H achieves 25.1% over prior SOTA on LawBench, 12.4% faster GPU kernels than prior SOTA (1,017 vs 1,161 {\mu}s), and 20.4% over prior SOTA on denoising. Harness updates make the model agentic, shaping how it searches and acts, while weight updates build the domain intuition that no prompt or scaffold can instil.
arXiv:2605.27176v2 Announce Type: replace
Abstract: Knowledge graphs (KGs) can provide structured scientific context to language models, but it remains unclear which graph facts actually shape the generated hypotheses. We study KG-guided hypothesis generation for battery materials across Mistral-7B, Llama-3.1-70B, and Gemini 2.5 Flash. We perturb local KGs by varying density, ontology richness, topology, and control structure, and evaluate outputs with both provided-graph and fixed-reference metrics. Across models, KG utility is selective and model-dependent: graph context changes outputs, but no-KG outputs also recover substantial graph content from model priors. Compact top-k subgraphs often approximate full-KG behavior, including when claimed-outcome triples are held out. At the same time, compression is not unique to one semantic ranking rule, random and topology-based subsets can also recover much of the signal. These results support a redundancy-aware Compressive KG hypothesis: useful KG signal is often recoverable from compact, scientifically structured subgraphs rather than requiring the full local graph.
arXiv:2605.27078v2 Announce Type: replace
Abstract: Training loss and accuracy are the standard signals used to monitor generalization during deep neural network training. Two well-documented phenomena complicate this picture: in grokking, train loss falls rapidly while test performance improves abruptly only after a long delay; in epoch-wise double descent, train loss decreases monotonically while test loss or error rises and falls. Existing accounts are often task-specific, and a task-agnostic analysis framework for diagnosing and explaining these phenomena across realistic tasks and architectures is missing. We address this challenge by analyzing two competing processes that underlie learning dynamics: representation learning in the encoder and readout calibration in the final classifier. Using tools from representational geometry, neural tangent kernels, and linear probing, we show that both processes are active throughout training, with the fluctuations of their relative speed giving rise to seemingly anomalous generalization dynamics. Applying the representation-readout decomposition to grokking across a wide range of tasks and architectures, we find that the readout is train-biased before grokking onset, and representation learning is gradual but not absent, contrary to the lazy-to-rich account. The framework further provides diagnostic signatures distinguishing spurious from genuine generalization: in a previously reported MNIST grokking example and an epoch-wise double descent example, apparent delayed or non-monotone generalization is shown to arise from representation degradation and readout misalignment induced by non-standard training recipes. Together, these results establish the representation-readout decomposition as a top-down framework for understanding learning dynamics and revealing underlying algorithms for interpretability research.