Weekly Recap

This lineage reframes traits as infrastructure-derived ecological observables, enabling trait-based ecology, conservation, and global change ecology, including climate-impact assessment, at scales commensurate with rapid environmental change.

A continental-scale synthesis of published plant–hummingbird interaction networks across the Americas is conducted to evaluate how well current data represent geographic, environmental, taxonomic, and trait dimensions of the Trochilidae and their plant partners and identifies priority gaps in species and regional coverage.

Decoy Direction Optimization (DDO), a fast, post-hoc weight-editing defense that requires no base-model finetuning, is introduced, based on a simple mechanistic insight: ablation attacks rely on contrastive estimators to find the refusal direction.

VC-Attention is proposed, a training-free low-bit attention framework that addresses diffusion Transformers and quantization scale by pairing Value smoothing with a fused probability Cast, and improves fidelity over low-bit baselines.

This study implements a hybrid-structure-based deep learning framework to generate monthly 2 m specific humidity, 2 m temperature and surface pressure at 1/30° × 1/30° horizontal resolution during 1901–2023, and reproduces learned terrain-related spatial gradients without explicitly enforcing physical equations or terrain constraints.

VC-Attention is proposed, a training-free low-bit attention framework that addresses diffusion Transformers and quantization scale by pairing Value smoothing with a fused probability Cast, and improves fidelity over low-bit baselines.

This study implements a hybrid-structure-based deep learning framework to generate monthly 2 m specific humidity, 2 m temperature and surface pressure at 1/30° × 1/30° horizontal resolution during 1901–2023, and reproduces learned terrain-related spatial gradients without explicitly enforcing physical equations or terrain constraints.

This study reconceptualizes pairwise classification not as a technique confined to offline RL, but as a domain-agnostic structural principle for determining which of the two candidates scores higher when a context is shared.

A continental-scale synthesis of published plant–hummingbird interaction networks across the Americas is conducted to evaluate how well current data represent geographic, environmental, taxonomic, and trait dimensions of the Trochilidae and their plant partners and identifies priority gaps in species and regional coverage.

Decoy Direction Optimization (DDO), a fast, post-hoc weight-editing defense that requires no base-model finetuning, is introduced, based on a simple mechanistic insight: ablation attacks rely on contrastive estimators to find the refusal direction.

This work introduces HarnessVLN, a zero-shot, training-free framework that unifies instruction-following and object-goal navigation through a shared Agent Harness, and demonstrates its applicability to both navigation tasks in real-world environments.

The Mind2Dialogue framework proposes a psychology-guided simulator that preserves personal characteristics while updating mental states through interaction to generate coherent conversations, and trains models on the Oracle's well-informed responses to assist users without direct access to their mental states at deployment.

This work shows that the frozen agent LLM already carries the routing signal in its own forward passes, and that two linear maps suffice to read it out with no skill text in the context, and shows that Gavel (Glance And Verdict from a frozen LLM) reads it in two steps.

ETHEREAL is presented, the first EV-GNN accelerator that scales to 640-times-480 resolutions, thanks to a neighbor-parallel spline convolution engine and a 2D/3D-split memory hierarchy with a novel region-of-interest spatiotemporal caching mechanism.

A dense, decision-level reward is introduced in which an LLM judge evaluates the necessity of each tool call, which effectively suppresses cue-driven tool use while preserving task performance, providing a practical approach to improving the robustness of LLM agent tool-use policies.

This work proposes ARIA (autoencoder-gated inference-time unlearning), a test-time unlearning method that leaves model weights intact and gates access to unwanted knowledge only when generation enters a forget-related state and introduces three post-unlearning adversarial attacks targeting weight-space and decoding-space recovery.

WMHA is exploited in WMHA, a latency-first diffusion-transformer inference accelerator: a very-long-instruction-word sequencer issues four engines from one instruction word, a weight-stationary 16x16 dual-dot array streams FP8 and BF16 contractions, and a single-pass online-softmax attention pipeline keeps keys and values resident through a skewed software pipeline.

WLA$^3$ (World Latent Action Modeling for Semantics, Dynamics, and Kinematics), a unified generalist policy model framework built around representations learned by a World Latent Action Model (WLAM), is introduced.

Flow-Matched Motion Priors (FMP) is introduced, an online scalar reward learned from paths connecting current rollout histories to an expert motion bank, and Controlled reward-model experiments show substantially better generalization beyond the fitting rollout than value-only or endpoint-only fitting.

This work defines the scaling inflection point as the per-session budget where marginal Elo gains match the independent-sampling reference, and proposes Elo-per-token analysis, which tracks the best solution found at each token budget and uses a Bradley-Terry model to aggregate within-task orderings into Elo ratings across tasks with different score scales.

This lineage reframes traits as infrastructure-derived ecological observables, enabling trait-based ecology, conservation, and global change ecology, including climate-impact assessment, at scales commensurate with rapid environmental change.

A mean-field analysis of attention, which shows how the average attention from one token to another defines a kernel that carries representations layer to layer and can be iterated through the network to model how the geometry is transformed.

CLEAR is proposed, an agentic framework for cross-source evidence adjudication in LLMs in medicine that independently generates candidate answers from three complementary pathways---parametric knowledge, locally curated corpora, and dynamically retrieved evidence---reflecting three common sources of information available to LLMs.

This work identifies a factual-salient layer span within LLMs whose derived signal is selectively elevated for factual tokens and exhibits anomalous spikes at hallucination-prone steps, and proposes DescaPE, a decoding framework that leverages internal model signals to suppress hallucination-prone trajectories at inference time.

This paper implements the white-box CLWE-RFF backdoor construction end to end using only numpy and scipy, to test whether this threat is realizable with commodity scientific-computing tools or requires specialized cryptographic infrastructure.

SWB-DM's cache carries a real warm-up cost, but extending all baselines to the same round budget shows its CIFAR-10 gains are disproportionately large, and on CIFAR-100, FLTrust benefits more -- for reasons entirely unrelated to caching.

This work examines exact-content binding, authorization freshness and checkpoint recovery in Lightcap's publication enforcement mechanism, and transfers retrieval models from ArguAna to SciFact and NFCorpus, and diagnostic decision rules from Thunderbird to BGL, distinguishing probability calibration from ranking changes.

Results show that ActGuard reduces attack success rates to a level comparable to state-of-the-art defenses while maintaining task utility close to the no-attack setting, achieving a favorable security-utility trade-off.

DR-SL (Dehydrate-Rehydrate with Self-Learning loop), which formalizes de-identification completeness as two measurable conditions: de-identification sufficiency under Pufferfish semantics, and task-information preservation via QA probes, is proposed.

A global sensitivity analysis of key LLM functions, which bounds the required scale of differentially private noise and derives an upper bound on floating-point error from masking and noise cancellation in the TEE as a function of the privacy parameter epsilon.

FairLint-DL, a Visual Studio Code extension that implements a shift-left approach to fairness testing by enabling pre-training, IDE-native bias detection directly on tabular datasets, demonstrates the feasibility of integrating fairness analysis into the developer workflow without significant overhead.

HazardAuditor is introduced, an execution-grounded framework that runs heterogeneous agents in controlled environments and normalizes their interactions into a canonical event representation for cross-framework supervision, and observes that token-level post-training objectives create a structural mismatch for generative guards, causing longer rationales to dominate gradient updates.

Qualitative examples show the PhysBrain 1.5 model's ability to produce end-effector trajectories and predict future scenes through spatially aligned RGB, depth, and robot-mask outputs.

RSIAgent is introduced, a training-free multi-agent framework for recursive self-improvement through autonomous memory construction that substantially improves strong open-source models, enabling Kimi-K3 and GLM-5.3 to outperform frontier closed-source models including GPT-6.3.

This work formalizes poison selection as oracle-budgeted set optimization and introduces SAILS (Set-level Audit-Informed Iterative Learned Selection), which learns a set scorer from a few hundred finetune-and-evaluate runs, ranks millions of candidate sets, and audits only a small shortlist.

This work identifies structured design specifications as a practical control point for exploring alternative UI concepts while keeping downstream generation settings fixed through an inference architecture that makes design direction an explicit intermediate decision.

This work forms a unified approach to capability formation and process-centered evaluation, enabling discovery behavior to be trained, improved, and measured beyond final-answer performance.

KaiNinja, a part-level extension of TRELLIS.2 built on a dual-volume form of its O-Voxel representation, is the first 3D generative model trained on agent-authored part data and finds that whole-object fidelity improves over the same backbone fine-tuned on the same dataset.

This work proposes an exploration-guided prompt scaffolding framework that adapts the training prompt distribution dynamically throughout RL post-training of multimodal large language models (MLLMs), and uses a lightweight rollout-based proxy for prompt utility derived from KL-regularized policy improvement theory.

ModaLens, a paired image-swap audit, measures how report availability changes image sensitivity and limits conclusions about visual correctness; the direction replicates in two further model lineages.

It is demonstrated that full-aggregate parallel suppression during from-scratch pretraining lowers validation-loss trajectories and improves downstream averages, with the value-space variant strongest.

The results show that the practical losslessness of Orthrus depends on numerical precision and that exact trajectory equivalence should be evaluated separately from downstream task performance.

Atria Dawn Preview is introduced, a foundation agentic language model designed for scientific research and engineering workflows, with the goal of expanding the frontier of agent productivity in the real world.

The Neverwhere Benchmark Suite is developed, comprised of over sixty 3D Gaussian Splatting reconstructions of urban indoor and outdoor scenes, to encourage large-scale and reproducible robot evaluation by making it easier to create and integrate Gaussian splats-based reconstructions into simulated continuous testing setups.

This work explores how diffusion can enable multilingual translation with a discrete diffusion framework that refines all target languages in parallel, achieving sublinear latency scaling with the number of targets, and supports deployment as a single unified model to replace multiple independent systems.

Applying SOTA models for Persian hate-speech detection on ParsHate shows moderate performance, especially with samples from earlier years, and low performance with target identification, which emphasizes the diverse sampling of hate speech in ParsHate and its challenging nature that requires more advanced methods for better performance.

Preliminary results reveal substantial differences in safety-utility calibration across models and show that failures can emerge only after several initially safe interaction steps, motivating treating agent safety as a trajectory-level property rather than a single-turn or binary success criterion.

Stellar Colosseum is introduced, a model-agnostic harness for allocating inference across research in mathematics and theoretical computer science that demonstrates the capabilities of Colosseum through open-ended research and evaluations on theorem-proving and competitive programming benchmarks.

The Deliberative Polling Diagnostic Framework is introduced, which compares human and LLM belief shifts after identical informational interventions and offers a concrete protocol: run the deliberative diagnostic before trusting LLM personas to mimic revised beliefs.

Overall, the results show that SAE- and probe-based techniques often outperform basic prompt-based methods for steering LLM agents, although this advantage depends on the specific prompting strategy involved.

The findings provide the first empirical evidence that seemingly"objective"physical descriptions can retain systematic gender associations in human interpretation, and uncover systematic patterns of model-human misalignment.

This work decomposes masked multi-query recall at a fixed state budget along three single-knob axes: a short causal convolution, the transition structure (rank-1 delta rule vs. diagonal), and decay.

CRN v2, a lightweight logit-level correction module that sits atop a fully frozen Gemma 4 E2B model that sits atop a fully frozen Gemma 4 E2B model is proposed, a study of a design principle (frozen base + logit correction + KL anchoring), not a claim of architectural novelty.

A matched ablation on unfiltered training data shows substantial benefits from filtering for all fine-tuned models, with the LLM-based models gaining roughly twice as much from filtering as the dedicated encoder-decoder MT system, supporting the view that for Belarusian MT one of the primary bottlenecks is data quality.

StalePO is introduced, an objective derived from three requirements this regime imposes that improves the fraction of segments passing all LLM-as-judge MQM quality checks by 14.9 and 4.6 percentage points, respectively, with gains concentrated on style and fluency.

Comparing matched records of the same cases under fixed labels and splits in three systems showed different patterns: the NHTSA defect summary remained strongest under every model family tested, whereas the ASRS analyst synopsis outperformed the reporter narrative under learned sequence models but not under lexical baselines.

This work builds a standardized harness over four clinical practice guidelines and evaluates twelve LLMs on 222 synthetic clinical questions, finding that most models can attach verbatim quotes to over 90% of their claims from prompting alone, apart from some lightweight models such as claude-haiku-4.5.

This method integrates the LMs directly into the parameters of an LLM-based ASR model, requiring no additional computational cost at inference, and consistently improved the ASR performance in the target domains, without degrading inference speed or memory footprint.

Results show that inoculation with a learned association introduced via midtraining can shape selective generalisation, and more work is needed before this approach can become a load-bearing component in a developer's safety framework.

This work proposes Learning to Coach (L2C), a framework that trains a dedicated LLM-as-a-Coach to extract actionable experiential knowledge from an actor model's previous trajectory, and studies two such rewards: a same-instance reward, which improves subsequent responses on the original problem, and a cross-instance reward, which elicits knowledge that transfers to other instances.

External evaluation shows that while citation validity remains strong, evidence utilization, span alignment, and refusal calibration become harder under domain shift, indicating that trustworthy RAG systems require explicit validation between retrieval and final answer delivery.

This work proposes an efficient method to add streaming ASR capabilities to an existing duplex S2S model by introducing a lightweight ASR head in parallel to the agent text head, enabling real-time user transcription while preserving full-duplex conversational capabilities including turn-taking and barge-in handling.

PeerPen, a writing assistance tool embedded within a Reddit-like interface, supporting two main features: draft generation and revision of user-written responses, is built and it is found that PeerPen reduced the burden of composing responses and increased confidence in offering support.

This work investigates parameter-efficient approaches for transferring knowledge from resource-rich source languages to low-resource target languages on Whisper, and proposes Sequential Adapter Stacking, which places a trainable target-language adapter on top of a frozen source-language adapter.

The historical-bounded large language models paradigm informs new forms of interactive experiments, where temporal knowledge boundaries become experimental variables, and expands the realm of science fiction science, which turns thought experiments into actual experiments.

This work introduces the CAST framework to connect measurement choices with person-specific models of exposure, behavior, physiology, and experience and proposes four synchronized measurement modules linking mobile and wearable data with self-reports and intervention responses.

This paper investigates how AI-driven feedback mechanisms, including sentiment analysis systems, pulse surveys, chatbots, engagement dashboards, and predictive analytics, are changing African workplaces through offering continuous listening, instant performance information and proactive engagement with employees.

This work proposes GPEvac: a GNN-based PPO framework that computes adaptive evacuation routes during shooting events, and shows that GPEvac outperforms intelligent baselines across distinct architectural layouts, significantly reducing total threat exposure.

This analysis finds substantial variation in coverage: AI risks related to robustness, system security, and governance receive more attention than socioeconomic, environmental, and emerging risks, including multi-agent risks.

Users retained, migrated, and reconstructed companions, finding that preserving conversation records did not necessarily restore shared memories or familiar interactions, and infrastructure dependence connected relational continuity and collective contestation during this national regulatory transition.

Overall, it is found that no model has a clear preference, nor a clear preference for a party, but that this depends on the specific issue or task the user asks about.

An in-the-wild study of task planning using time-stamped logs of 24,265 tasks from 957 users of a widely used daily planner app, observed over six weeks alongside self-reported surveys to support evaluating planning tools across the task lifecycle.

Energy transition decision-making is pervaded by deep uncertainties, and computer models can give insight into techno-economic complexity but alone are insufficient, as part of the uncertainties involve justice dilemmas.

This article presents a fully static, whole-program binary lifting system requiring no runtime translation support on the target, and implements this approach as a prototype binary recompiler from x86-64 binaries to LLVM IR, requiring no code/data heuristics.

RuleAutoPilot is presented, an end-to-end agentic framework that generates deployable Suricata rules directly from malware network traffic, with no prior threat intelligence required, with a Benign Traffic Fingerprinting stage that removes known benign background flows before LLM processing.

RAPID is proposed, a self-referenced latent maximization framework that removes external dependencies and encourages the same model update to induce consistently disruptive effects across training samples, thereby improving reachability and establishing a paradigm for real-time protection against unauthorized distillation in deployed T2I systems.

Market context improves action and ticker prediction in the controlled ablation, while transaction sizing remains difficult, and models overproduce hold actions, underpredict sell decisions, and simplify multi-security transactions.

This work develops a Carbon-Aware AI Procurement Framework (CAAPF), a Green IS design artifact that operationalizes sustainable AI governance for enterprise procurement, and suggests that environmental responsibility and decision quality can be mutually reinforcing.

Multi-factor authentication platforms that combine multiple mechanisms that combine multiple mechanisms focus on strategies that prevent replay attacks (i.e. the reuse of a component that could have been potentially compromised).

This paper demonstrates the full potential of behavior-preserving symmetries as a defense against stegomalware, as well as the risks these symmetries pose when exploited by attackers, and quantifies the loss in model performance associated with applying these methods.

This paper explores how to apply formal verification to artificial organs, stage by stage, and what each technique actually guarantees about the device.

This work reveals an intrinsic vulnerability in stereo cameras that stems from their pixel sampling and calibration processes, which can influence the outputs of stereo matching algorithms and proposes a novel strategy that leverages similarity scores to dynamically detect and suppress the depth discrepancies.

This paper derives general lower and upper bounds on the PIR capacity for general graphs, under certain privacy requirements, that capture the essence of both LPIR and classical PIR.

A schema-aware split learning (SL) framework that preserves privacy, using a large language model (LLM) as a shared semantic encoder to harmonize heterogeneous survey schemas across institutions and enables accurate, privacy-preserving, and resource-efficient collaborative learning from heterogeneous mental health survey data.

A snapshot software receiver implementing a simplified version of the Galileo SAS protocol is developed to compute the receiver clock bias from an authenticated pseudorange, using radio-frequency data recorded with an engineering prototype software-defined radio receiver from Septentrio.

A co-design study where 14 participants redesigned Google and Apple Health interfaces using Figma Make, finding conversational AI designs converged around chat-window conventions and implications for preserving agency and flexible user-directed interfaces.

The biomedical significance is that, under the stated threat model, a served genomic model can process an encoded sequence without exposing plaintext queryderived activations to the compute provider.

A principal hierarchy spanning human user, operator/deployer, orchestrator agent, sub-agent, and tool endpoint as an organizing framework is introduced, and seven structural requirements are proposed, derived from a four-layer reference architecture, and runtime enforcement and aggregation bounds are identified as the principal unresolved gaps.

The results support a narrow conclusion: the verifier changes what the system ships, while deterministic code supplies enforcement and auditability; they do not establish superiority to other agents or generalization beyond lab targets.

An event-native symbolic-temporal spike encoding framework that maps heterogeneous cyber events directly into sparse, spike-compatible inputs is introduced, assigning encoding roles to semantic identity, local frequency context, and inter-event timing, which preserves categorical semantics and temporal dynamics.

Two universal tool-based defenses are introduced: Attacker Tool Filtering, which uses anomaly detection to identify and remove suspicious tools, and Normal Tool Recalling, a white-box method that restores the agent's original toolset prior to planning.

DescGuard is proposed, a registration-time defense that restores the targeted planning metrics and downstream performance toward their baseline levels without modifying worker implementations, the planner, or the orchestration logic, and composes with existing isolation, permission-control, and runtime mechanisms.

QPriv-VL is proposed, a question-guided, privacy-aware token-pruning framework for FL, SL, and USL that prunes visual tokens before transmission based on task utility and privacy sensitivity and evaluates it on GQA, OK-VQA, VQAv2, SLAKE, VQA-RAD, and PathVQA against four privacy attack families.

ZEBRA is a fully automated verification and bug-detection framework that reduces zkVM verification to a solution-set cardinality problem over a canonical trace space, where redundancies such as null-row padding and non-deterministic permutations are eliminated prior to counting.

PIDS-Bench is presented, a frozen multi-axis benchmark that jointly evaluates attack detection and benign false-positive behavior at fixed thresholds, spanning in-distribution inputs, hard-benign prompts that mimic injection structure without malicious intent, obfuscated attacks, and domain and structural distribution shifts.