arXiv:2603.19828v3 Announce Type: replace
Abstract: Autoformalization aims to produce formal statements that compile and faithfully preserve the intended meaning of informal mathematics. Yet standard single-output evaluation protocols collapse a many-to-many problem into a single-output prediction task. For downstream proving, this granularity is too coarse: a formal statement is not merely a faithful translation endpoint, but also a prover-facing interface whose structure can alter proof search under a fixed budget. We therefore recast autoformalization as budgeted test-time search: FormalEvolve maintains a compilation-feasible archive for reuse, while reporting the deduplicated semantically accepted repertoire for evaluation and downstream proving. It expands the archive with LLM-driven mutation, crossover, bounded patch repair, and symbolic Abstract Syntax Tree (AST) rewrites for structural diversity. Under a generator-call budget of T=100 with a fixed LLM semantic judge, FormalEvolve reaches SH@100 of 58.0% on CombiBench and 84.9% on ProofNet, improving over all no-archive controls while reducing the cross-problem concentration of semantic successes. To assess downstream value, we evaluate the resulting repertoires under a fixed B=64 prover budget, where they improve theorem-complete proving over the matched no-archive control; additional stronger-base statement-generation experiments show that archive-search gains hold with stronger seed and repair models. Manual faithfulness audits calibrate these judge-positive outputs.
Science Journals
arXiv:2603.23234v2 Announce Type: replace
Abstract: LLM agents increasingly rely on memory mechanisms to reuse knowledge from past problem-solving experiences. However, existing methods typically construct memory for a single agent and reuse it with the same underlying model, tightly coupling stored knowledge to model-specific reasoning styles. In heterogeneous deployments, where agents may be instantiated with backbone models of different sizes, architectures, or specializations, this raises a key question: can a single memory system be shared across agents with different backbone models? We find that naive cross-model memory transfer can degrade performance, because stored memories often entangle task-relevant knowledge with model-specific biases. To address this challenge, we propose MemCollab, a collaborative memory framework that builds shared cross-model memory by contrasting reasoning trajectories generated by different model-based agents on the same task. Through this contrastive process, MemCollab distills abstract reasoning constraints that capture shared task-level invariants while suppressing model-specific artifacts. We further introduce a task-aware retrieval mechanism that conditions memory access on task category, ensuring that only relevant constraints are retrieved at inference time. Experiments on mathematical reasoning and code generation benchmarks show that MemCollab consistently improves both accuracy and inference-time efficiency across diverse agents, including settings with different model families. These results demonstrate that collaboratively constructed cross-model memory can serve as a shared reasoning resource for heterogeneous LLM-based agents.
arXiv:2603.26668v2 Announce Type: replace
Abstract: As an important paradigm for enhancing the generation quality of Large Language Models (LLMs), retrieval-augmented generation (RAG) faces the two challenges regarding retrieval accuracy and computational efficiency. This paper presents a novel RAG framework called Bridge-RAG. To overcome the accuracy challenge, we introduce the concept of abstract to bridge query entities and document chunks, providing robust semantic understanding. We organize the abstracts into a tree structure and design a multi-level retrieval strategy to ensure the inclusion of sufficient contextual information. While this hierarchical organization substantially improves answer quality, traversing the tree to locate the abstracts that contain a query entity inevitably introduces additional retrieval overhead. To restore retrieval efficiency, we further integrate the Cuckoo Filter in CFT-RAG, which provides O(1) entity lookup and naturally fits the entity-to-abstract pathway of our framework. Extensive experiments show that Bridge-RAG achieves consistent accuracy improvements across all metrics and up to $1.9\times$ faster retrieval compared to structured RAG baselines.
arXiv:2603.28949v2 Announce Type: replace
Abstract: The artificial intelligence industry is not an isolated economic phenomenon; it is the current physical substrate for a broader, multi-billion-year process: the evolution of an abstract intelligence on Earth. As the scale of computation accelerates planetary-wide, the academic and popular discourse must address the physics aspect of this phase transition, beyond algorithmic architectures, alignment, and silicon supply chains. From first principles, it is clear that if the current exponential trajectory of computation holds, the ultimate bottleneck of the coming decades will be neither data nor capital, but the laws of thermodynamics and the finite heat capacity of the Earth. The evolution of intelligence is fundamentally a problem of non-equilibrium thermodynamics, bound by strict hardware limitations, and ultimately, an absolute ecological boundary condition. Civilization itself is an exceedingly rare and highly expensive thermodynamic algorithm actively fighting the default settings of the cosmos. To better understand AI's future and our future, we must rigorously examine the physical laws governing computation, complexity theory, and the narrow thermodynamic tightrope our civilization must walk to survive it
arXiv:2605.30131v1 Announce Type: new
Abstract: Radiology report generation (RRG) is commonly formulated as a single-path generation task, where a multimodal large language model (MLLM) produces one decoded report as the final output. While recent progress has largely been driven by scaling training data, model capacity, and retrieval mechanisms, improving report quality at inference time remains underexplored. In this work, we observe that fixed radiology MLLMs often generate clinically stronger reports elsewhere in their candidate pool than the one selected by default decoding, suggesting that inference-time decision making remains an overlooked bottleneck. To address this, we propose Clinical Consensus Selection (CCS), a decoder-agnostic inference-time selection framework that samples multiple candidate reports and selects the one with the highest clinical consensus across the rollout pool. CCS unifies text-based utilities with a radiology-adapted utility computed by an image--report-trained multimodal embedder, which measures candidate agreement beyond surface-level textual similarity. Across three datasets and multiple radiology MLLMs, CCS consistently improves inference-time performance over single-path decoding and generic Best-of-N baselines, with particularly clear gains on clinical metrics. Further analysis shows that image-grounded utility forms a selection axis distinct from textual consensus and that substantial headroom remains for improving RRG at inference time.
arXiv:2605.30126v1 Announce Type: new
Abstract: Large Vision-Language Models (LVLMs) map visual inputs into dense token sequences, imposing a quadratic computational bottleneck for inference. Elastic visual-token compression addresses this by training a single model that can run at multiple visual-token budgets. However, existing approaches struggle under aggressive compression. Spatial-only compression, as in nested pooling, behaves as an imperfect low-pass filter and induces spectral aliasing that obscures fine-grained detail. Query-only compression, as in nested query resampling, replaces explicit grid-aligned tokens with non-local summaries and substantially degrades spatial grounding. To resolve this representational conflict, we introduce PARCEL (Pool-Anchored Resampling with Conditioned Elastic Queries for Efficient Vision-Language Understanding), a visual tokenization architecture that dynamically partitions the labor of feature extraction. PARCEL establishes spatial pool tokens as low-frequency layout anchors and conditions elastic query tokens on these anchors through Pool-Conditioned Query Resampling. This encourages query tokens to focus on complementary visual features rather than redundant spatial mapping. Extensive evaluations across 27 benchmarks show that PARCEL improves the performance-efficiency Pareto frontier, consistently outperforming existing matryoshka baselines across visual-token budgets while preserving the "train once, deploy anywhere" paradigm.
arXiv:2605.29099v1 Announce Type: new
Abstract: As server CPUs scale to dozens and now hundreds of cores per socket, parallel query engines must rethink how they redistribute data between threads. Partitioned operators such as hash joins and aggregations require frequent data redistribution across threads, yet existing intra-process shuffle designs fundamentally fail to scale with core count: batch partitioning avoids cross-thread synchronization in the hot path but materializes all intermediate data, introduces a global producer/consumer barrier, and requires a consumption approach with low cache locality, while channel-based streaming avoids materialization but incurs per-channel synchronization that scales poorly with core count. As core counts rise, these architectural tradeoffs increasingly prevent engines from fully utilizing modern hardware.
We present a ring-buffer streaming shuffle design that addresses these shortcomings through lock-free atomic slot acquisition into fixed-size batch groups, achieving amortized O(1) synchronization cost per batch and O(M) memory independent of input size. Ring-buffer shuffle has been implemented in Redpanda's Oxla query engine for two years, where it currently powers production queries for Redpanda SQL users.
We evaluate all three approaches on a 72-core NVIDIA GraceHopper, a 192-core dual-socket AWS Graviton4, and a 96-core (192-thread) AMD EPYC. On a 72-core single-socket system the ring buffer outperforms channel streaming by up to 44% and batch partitioning by up to 79%; at 192 cores the advantage over channel grows to over 100% and over 300% versus batch partitioning. Even so, on chiplet architectures with many partitioned L3 caches, the shared atomic counter becomes a cross-die bottleneck and channel-based streaming remains competitive. End-to-end Graviton4 evaluation on TPC-H (21 queries) and ClickBench (43 queries) shows the advantage is workload-shape-dependent.
arXiv:2604.04956v3 Announce Type: replace
Abstract: The recent, super-exponential scaling of autonomous Large Language Model (LLM) agents signals a broader, fundamental paradigm shift from machines primarily replacing the human hands (manual labor and mechanical processing) to machines delegating for the human minds (cognition, reasoning, and intention). The uncontrolled offloading and scaling of "thinking" itself, beyond human's limited but efficient biological capacity, has profound consequences for humanity's heat balance sheet, since thinking, or intelligence, carries thermodynamic consequences. The Earth has already surpassed the heat dissipation threshold required for long-term ecological stability, and projecting based on empirical data reveal a concerning trajectory: without radical structural intervention, anthropogenic heat accumulation will breach critical planetary ecological thresholds in less than 6.5 years, even under the most ideal scenario where Earth Energy Imbalance (EEI) holds constant. In this work, we identify six factors from artificial intelligence that influence the global heat dissipation rate and delineate how their interplay drives society toward one of four broad macroscopic trajectories. We propose that the integration of artificial intelligence and its heat dissipation into the planetary system constitute the tenth planetary boundary (9+1). The core empirical measurement of this boundary is the net-new waste heat generated by exponential AI growth, balanced against its impact on reducing economic and societal inefficiencies and thus baseline anthropogenic waste heat emissions. We demonstrate that managing AI scaling lacks a moderate middle ground: it will either accelerate the breach of critical planetary thermodynamic thresholds, or it will serve as the single most effective lever on stabilizing the other nine planetary boundaries and through which safeguarding human civilization's survival.
Cognitive Loop of Thought: Reversible Hierarchical Markov Chain for Efficient Mathematical Reasoning
arXiv:2604.06805v2 Announce Type: replace
Abstract: Multi-step Chain-of-Thought (CoT) has significantly advanced the mathematical reasoning capabilities of LLMs by leveraging explicit reasoning steps. However, the widespread adoption of Long CoT often results in sequence lengths that exceed manageable computational limits. While existing approaches attempt to alleviate this by reducing KV Cache redundancy via Markov chain-like structures, they introduce two critical limitations: inherent memorylessness (loss of context) and limited backward reasoning capability. To address these limitations, we propose a novel Chain-of-Thought framework based on Reversible Hierarchical Markov Chain, termed Cognitive Loop of Thought (CLoT), and a backward reasoning dataset CLoT-Instruct. In CLoT, problems are decomposed into sub-problems with hierarchical dependencies. Inspired by human cognitive processes, we introduce a backward verification mechanism at each hierarchical layer. Furthermore, we implement a pruning strategy: once higher-level sub-problems are verified, redundant lower-level sub-problems are pruned to maximize efficiency. This approach effectively mitigates error propagation and enhances reasoning robustness. Experiments on four mathematical benchmarks demonstrate the effectiveness of our method. Notably, on the AddSub dataset using GPT-4o-mini, CLoT achieves 99.0% accuracy, outperforming traditional CoT and CoT-SC by 4.1% and 2.9%, respectively.
arXiv:2604.09557v2 Announce Type: replace
Abstract: Speculative Decoding (SD) has emerged as a critical technique for accelerating Large Language Model (LLM) inference. Unlike deterministic system optimizations, SD performance is inherently data-dependent, meaning that diverse and representative workloads are essential for accurately measuring its effectiveness. Existing benchmarks suffer from limited task diversity, inadequate support for throughput-oriented evaluation, and a reliance on high-level implementations that fail to reflect production environments. To address this, we introduce SPEED-Bench, a comprehensive suite designed to standardize SD evaluation across diverse semantic domains and realistic serving regimes. SPEED-Bench offers a carefully curated Qualitative data split, selected by prioritizing semantic diversity across the data samples. Additionally, it includes a Throughput data split, allowing speedup evaluation across a range of concurrencies, from latency-sensitive low-batch settings to throughput-oriented high-load scenarios. By integrating with production engines like vLLM and TensorRT-LLM, SPEED-Bench allows practitioners to analyze system behaviors often masked by other benchmarks. We highlight this by quantifying how synthetic inputs overestimate real-world throughput, identifying batch-size dependent optimal draft lengths and biases in low-diversity data, and analyzing the caveats of vocabulary pruning in state-of-the-art drafters. We release SPEED-Bench to establish a unified evaluation standard for practical comparisons of SD algorithms.
arXiv:2605.01194v2 Announce Type: replace
Abstract: Vision-Language-Action (VLA) models have demonstrated remarkable capabilities and generalization in embodied manipulation. However, their decision-making relies on a fast, instinctive process that lacks deliberation. This strategy often leads to suboptimal or catastrophic actions when facing complex or ambiguous scenarios that require greater consideration. In this paper, we introduce \textbf{VLA-ATTC}, a framework that endows VLA models with adaptive test-time compute (TTC). VLA-ATTC employs an uncertainty-based ``cognitive clutch'' to dynamically transition from reflexive execution to a TTC deliberation phase when necessary. During TTC phase, a novel \textbf{Relative Action Critic} (RAC) model identifies the optimal action from generated candidates via pairwise comparisons. This relative mechanism replaces unstable absolute value estimation, significantly simplifying the learning objective. Furthermore, we introduce an efficient sampling strategy to amortize computational costs and an automated data pipeline that curates preference pairs without manual annotation. On the LIBERO-LONG benchmark, VLA-ATTC reduces the failure rate of the SOTA model PI0.5 by over 50\%. We will open-source all the code and weights.
arXiv:2604.13197v2 Announce Type: replace
Abstract: Process reward models (PRMs) provide fine-grained supervision for reasoning, but reliable PRMs often require step annotations or heavy verification pipelines, making them costly to scale and refresh during online RL. Implicit PRMs reduce this cost by training log-likelihood-ratio rewards from trajectory-level outcome labels. However, the log-ratio is constrained only as a sequence-level aggregate during training, while inference decomposes it into token- or step-level scores for partial prefixes. This train-inference mismatch leaves local credits weakly identified, so distribution-wide scoring can amplify misleading advantages. We propose Implicit Prefix-Value Reward Model (IPVRM), which directly learns the probability of eventual correctness for each prefix from outcome labels. Step signals are then obtained as temporal-difference (TD) differences between consecutive prefix values, aligning the training target with inference-time use. IPVRM markedly improves step-verification F1 on ProcessBench. To exploit these prefix values during policy optimization, we further introduce Distribution-Level RL (DistRL), which applies TD advantages to both sampled tokens and high-probability candidate tokens, providing dense counterfactual updates without additional rollouts. Experiments show that DistRL brings limited gains with unreliable implicit rewards, but consistently improves downstream reasoning when paired with IPVRM. The implementation of our method is available at https://github.com/gaoshiping/IPVRM .
arXiv:2605.30117v1 Announce Type: new
Abstract: Understanding how Vision-Language-Action (VLA) models transform multimodal knowledge into embodied control remains an open challenge. We present VLA-Trace, a progressive diagnostic framework that analyzes VLA models through a unified evidence chain from representation dynamics to causal control attribution and behavioral manifestation. It specifically combines cross-modal and checkpoint-drift centered kernel alignment (CKA) to trace representation evolution, attention knockout interventions to identify modality-specific control pathways, and rollout-level behavioral probes to examine grounding, shortcut dependence, and semantic following. Experiments on $\pi_{0.5}$ and OpenVLA reveal three key findings. First, the two models exhibit distinct modality-specific adaptation dynamics during VLA finetuning. Second, they rely on different multimodal routing strategies and layer-wise dependencies during action decoding. Third, although VLA policies excel at visually grounded trajectory generation, they remain limited in fine-grained semantic following. These findings highlight future directions for representation-preserving adaptation, causal VLA circuits, and compositional semantic control.
arXiv:2604.13519v2 Announce Type: replace
Abstract: Tool calling has greatly expanded the practical utility of large language models (LLMs) by enabling them to interact with external applications. As LLM capabilities advance, effective tool use increasingly involves multi-step, multi-turn interactions to solve complex tasks. However, the resulting growth in tool interactions incurs substantial latency, posing a key challenge for real-time LLM serving. Through empirical analysis, we find that tool-calling traces are highly structured, conform to constrained schemas, and often exhibit recurring invocation patterns. Motivated by this, we propose ToolSpec, a schema-aware, retrieval-augmented speculative decoding method for accelerating tool calling. ToolSpec exploits predefined tool schemas to generate accurate drafts, using a finite-state machine to alternate between deterministic schema token filling and speculative generation for variable fields. In addition, ToolSpec retrieves similar historical tool invocations and reuses them as drafts to further improve efficiency. ToolSpec presents a plug-and-play solution that can be seamlessly integrated into existing LLM workflows. Experiments across multiple benchmarks demonstrate that ToolSpec achieves up to a 4.2x speedup, substantially outperforming existing training-free speculative decoding methods.
arXiv:2604.15698v4 Announce Type: replace
Abstract: We study lossy compression of a finite statement source generated in a fixed deductive environment. The source symbols are statements in a knowledge base endowed with a shared proof system, and reconstruction fidelity is measured by preservation of deductive closure rather than by symbolwise equality. Fixing the proof system and a canonical scan order yields a decomposition of the source alphabet into an irredundant core and redundant stored consequences. At zero distortion, each core symbol induces a set of distortion-free reconstructions. In the nonconfusable (disjoint-core) regime, we show that the minimum zero-distortion rate equals the source mass of the core times the entropy of the source conditioned on that core. In the general confusable-core regime, we characterise the exact zero-distortion rate via a hypergraph-entropy quantity induced by jointly realisable core subsets, with a reduction to Korner-style graph entropy under a natural pairwise realisability condition. For reconstruction alphabets contained in the deductive closure of the source knowledge base, we further prove that the full rate-distortion function depends only on the core, so redundant states are invisible to both rate and distortion. Finally, when the decoder is limited to a bounded inference-depth budget (a bounded number of iterations of the immediate-consequence operator), we obtain an exact rate-depth-distortion characterisation. Under an additional order-robustness assumption identifying the chosen core with the order-free essential set, this characterisation interpolates between classical symbolwise compression and unconstrained deductive compression.
arXiv:2605.30115v1 Announce Type: new
Abstract: This work presents the Large Depth Completion Model (LDCM), a simple, effective, and robust framework for single-view metric depth estimation with sparse observations. Without relying on complex architectural designs, LDCM generates metric-accurate dense depth maps using a transformer. It outperforms existing approaches across diverse datasets and sparse observations. We achieve this from two key perspectives: (1) leveraging existing monocular foundation models to improve the quality of sparse depth inputs, and (2) reformulating training objectives to better capture geometric structure and metric consistency. Specifically, a Poisson-based depth initialization strategy is first introduced to generate a uniform coarse dense depth map from diverse sparse observations, providing a strong structural prior for the network. Regarding the training objective, we replace the conventional depth head with a point map head that regresses per-pixel 3D coordinates in camera space, enabling the model to directly learn the underlying 3D scene structure instead of performing pixel-wise depth map restoration. Moreover, this design eliminates the need for camera intrinsic parameters, allowing LDCM to naturally produce metric-scaled 3D point maps. Extensive experiments demonstrate that LDCM consistently outperforms state-of-the-art methods across multiple benchmarks and varying sparsity levels in both depth completion and point map estimation, showcasing its effectiveness and strong generalization to unseen data distributions.
arXiv:2605.22080v2 Announce Type: replace
Abstract: We introduce JMed48k, a multi-profession Japanese healthcare licensing benchmark for evaluating vision-language models. Built from official PDF materials released by the Japanese Ministry of Health, Labour and Welfare, JMed48k contains 48,862 exam questions and 20,142 images from 11 national licensing examinations between 2005 and 2025, with visual content annotated under an 8-type taxonomy. From this corpus, we derive JMed48k-Eval, a recent five-year evaluation subset with 12,484 scored questions, including 9,905 text-only questions and 2,579 questions with images. We evaluate 21 proprietary, open-source, and medical-specific models, reporting text-only and with-image performance separately. Because these subsets contain different questions, we further introduce a paired image-removal audit that evaluates questions with images before and after removing visual content to explore four answer-transition states. The audit shows that proprietary and open source models gain substantially from images, whereas medical-specific systems show limited observable use of visual evidence, with many correct answers persisting after image removal. Even among proprietary models, the net image-removal effect varies sevenfold across professions, from +5.7 points on Physician questions to +39.8 points on Public Health Nurse questions. We release JMed48k to support reproducible, profession-stratified evaluation of vision-language models in medical licensing settings.
arXiv:2605.30100v1 Announce Type: new
Abstract: World models require state tracking, which is the ability to maintain a correct latent state across action sequences. Existing benchmarks are often synthetic or language-based, limiting their value as tests of structured state updates in realistic domains. We introduce Chess-World-Model, a large-scale state-tracking benchmark built from 10 million real chess games, where models predict the exact board state reached after a sequence of legal moves. Alongside a held-out real-game split, we include an out-of-distribution split from uniformly random legal play, which tests whether models learn the transition rules rather than shortcuts from common human positions. Prior theoretical and empirical work has shown that Transformers struggle to state-track, while input-dependent linear RNNs require expressive state-transition matrices to do so. We therefore benchmark a causal Transformer, block-diagonal SLiCE, Mamba-3, and Gated DeltaNet with negative eigenvalues under a matched interface and training protocol. The recurrent models strongly outperform the Transformer at 3 and 8 million parameters. Real-game performance saturates above 18 million parameters, but the random-uniform split remains discriminative up to 40 million, exposing failures otherwise hidden by scale. Additionally, ablations show that less expressive state-transition mechanisms reduce performance on the out-of-distribution split for all three recurrent models. Together, these results establish Chess-World-Model as a practical large-scale benchmark for state tracking that exposes failures model scale would otherwise conceal.
arXiv:2605.30093v1 Announce Type: new
Abstract: Foundation features from self-supervised vision models and text-to-image diffusion models have proven effective for semantic correspondence estimation. However, because these features are learned primarily from 2D image objectives, they lack explicit 3D awareness and often confuse symmetric object sides, repeated parts, and visually similar structures that are distinct in 3D. We introduce a 3D-aware post-training framework that goes beyond available 2D foundation features by incorporating priors from 3D foundation models. Given an image, our method uses SAM3D to estimate object geometry and pose, and refines the pose through render-and-compare optimization. Subsequently, we render PartField descriptors from the reconstructed geometry into the image plane based on the estimated object pose. The resulting geometry-aware feature maps complement DINO and Stable Diffusion features, while geodesic distances on the reconstructed shapes enable reliable filtering of candidate correspondences. We use the filtered matches as supervision to train a lightweight adapter on top of DINO and Stable Diffusion for semantic correspondence. In contrast to prior post-training approaches that require pose annotations and rely on coarse spherical geometry, our method automatically obtains instance-specific 3D structure and uses it to guide correspondence learning. Experiments show that our approach improves semantic correspondence over the prior methods while reducing manual geometric supervision. Code and model can be found at https:/github.com/GenIntel/3D-SC.
arXiv:2605.30080v1 Announce Type: new
Abstract: Tokenization-free hierarchical models are emerging as a promising alternative to traditional Large Language Models (LLMs), addressing inherent preprocessing issues such as vocabulary design complexity, out-of-vocabulary (OOV) errors, and language-specific constraints. However, a significant challenge in these byte-level methods is the optimization of the compression ratio, a critical factor that dictates model performance for processing bytes data via chunks. In this paper, we propose Adaptive Targeted Dynamic Chunking (ATDC), a novel byte-compression control mechanism designed to enhance the effectiveness of dynamic chunking within hierarchical architectures. Our approach utilizes curriculum learning to progressively adjust the compression ratio during training, transitioning from low to high compression to stabilize the learning process. We provide an analysis establishing the relationship between the target compression ratio and Bytes-Per-Innermost-Chunk (BPIC), allowing for tracking of chunk-size evolution throughout the training phase. Evaluations conducted on the FineWeb-Edu 100B dataset demonstrate that hierarchical models equipped with ATDC achieve competitive Bits-Per-Byte (BPB) performance compared to conventional baselines operating at both byte and token levels. Furthermore, the proposed method exhibits more stable training dynamics and superior final performance across diverse downstream tasks compared to models using fixed compression ratios, while maintaining the inherent robustness and flexibility of byte-level processing.
arXiv:2605.30076v1 Announce Type: new
Abstract: Activation-based control steers large language models (LLMs) by intervening on their internal representations during inference, and has emerged as an effective paradigm for controlling behaviors such as persona and style. However, existing methods often rely on fixed steering directions or task-specific intervention modules, making them difficult to adapt to fine-grained concepts and compositional constraints. We propose UniSteer, a text-guided activation flow matching model that learns a conditional distribution over residual-stream activations from natural-language conditions. Instead of fitting a separate intervention for each target behavior, UniSteer learns a universal conditional velocity field in activation space. At inference time, UniSteer performs flow inversion by partially transporting a source activation toward a latent state and regenerating it under a target textual condition before injecting it back into the frozen LLM. The same conditional model supports activation-space classification by selecting the textual label with the lowest reconstruction energy. Experiments on three target LLMs show that UniSteer provides a unified interface across behavioral control, truthfulness steering, fine-grained concept steering, multi-constraint instruction following, and activation-space classification.
arXiv:2605.13841v2 Announce Type: replace
Abstract: Voice agents, artificial intelligence systems that conduct spoken conversations to complete tasks, are increasingly deployed across enterprise applications. However, no existing benchmark jointly addresses two core evaluation challenges: generating realistic simulated conversations, and measuring quality across the full scope of voice-specific failure modes. We present EVA-Bench, an end-to-end evaluation framework that addresses both. On the simulation side, EVA-Bench orchestrates bot-to-bot audio conversations over dynamic multi-turn dialogues, with automatic simulation validation that detects user simulator error and appropriately regenerates conversations before scoring. On the measurement side, EVA-Bench introduces two composite metrics: EVA-A (Accuracy), capturing task completion, faithfulness, and audio-level speech fidelity; and EVA-X (Experience), capturing conversation progression, spoken conciseness, and turn-taking timing. Both metrics apply to all major agent architectures, enabling direct cross-architecture comparison. EVA-Bench includes 213 scenarios across three enterprise domains, a controlled perturbation suite for accent and noise robustness, and pass@1, pass@k, pass^k measurements that distinguish peak from reliable capability. Across 12 systems spanning all three architectures, we find: (1) no system simultaneously exceeds 0.5 on both EVA-A pass@1 and EVA-X pass@1; (2) peak and reliable performance diverge substantially (median pass@k--pass^k gap of 0.44 on EVA-A); and (3) accent and noise perturbations expose substantial robustness gaps, with effects varying across architectures, systems, and metrics (mean $\Delta$ up to 0.314). We release the full framework, evaluation suite, and benchmark data under an open-source license.
arXiv:2605.29093v1 Announce Type: new
Abstract: In cloud data platforms, developers often encounter performance regressions that occur in specific tenant datasets. However, due to confidentiality constraints, they cannot access the original data, which makes it difficult to reproduce these regressions locally. Current methods for synthetic data usually focus on statistical properties, such as matching data distributions or improving query accuracy. However, they overlook the physical properties that control how the engine behaves during scans, including row-group pruning.
We propose ScanTwin, a lightweight framework that extracts a per-row-group sketch from the Parquet footer, including boundary values and compressed sizes, and releases them under $\varepsilon$-differential privacy using a boundary parameterization. On TPC-H and SSB (6M rows), ScanTwin achieves 0% pruning error and less than 1% byte error at $\varepsilon{=}\infty$. Under $\varepsilon{=}5$, high-selectivity queries ($>$30%) incur below 8.5% pruning error on both datasets, and per-query scan timing on DuckDB closely tracks the original.
arXiv:2605.30024v1 Announce Type: new
Abstract: Standard EU energy system modelling approaches optimize for least-cost, leading to highly centralized systems, in conflict with political feasibility and physical security concerns. This paper incorporates decentralisation as a constraint in a European energy system model using a novel, linear load-weighted renewable capacity constraint, the K-parameter, which scales with total system renewable capacity to avoid interference with decarbonisation targets. The model is a 37-node electricity-only brownfield system based on the PyPSA-EUR framework, with projected 2050 loads and technology costs. A total of 105 optimized scenarios are analyzed at 14 levels of decarbonization and 8 levels of decentralization. Full decarbonization leads to an 80% cost increase due to, among other factors, a 78% increase in energy generation capacity. Without decentralisation constraints, system equity initially improves but collapses at high decarbonisation levels due to concentration in regions with optimal renewable resources. Moderate decentralization of K=7 achieves 76% of the equity benefits at only a 9% cost increase compared to K=1. This indicates that moderate decentralization can be a viable strategy to balance societal preferences and cost-efficiency in the European energy transition.
arXiv:2605.30014v1 Announce Type: new
Abstract: Urban trajectories play a crucial role in modeling urban dynamics and supporting various smart city applications. However, privacy concerns restrict access to large-scale and high-quality trajectory datasets. Trajectory generation provides a promising alternative by synthesizing realistic data to mitigate privacy risks. However, existing methods fail to explicitly capture travel patterns and can only generate fixed-length trajectories under a single condition. To address these limitations, we propose \textbf{HTP}, which \textbf{H}ierarchically generates \textbf{T}ravel patterns first and then generates GPS \textbf{P}oints by using large language models (LLMs), rather than directly generating GPS points. We first design a trajectory-specific residual quantization variational autoencoder (RQ-VAE) that quantizes micro-level GPS trajectories into compact, macro-level travel pattern tokens in a coarse-to-fine manner. These tokens capture rich segment spatial irregularities, such as point density variations caused by traffic conditions. Then, we extend the LLM vocabulary with travel pattern tokens to align trajectory representations with the LLM input, and apply supervised fine-tuning (SFT) to align the LLM with the trajectory generation task, enabling generation of travel pattern sequences under various conditions. Extensive experiments on two real-world datasets show that HTP outperforms the strongest baseline by an average of 29.78\% in terms of generation quality. Our code is available at https://github.com/slzhou-xy/HTP.