Academic Intelligence · Curated Daily

探索全球前沿学术脉络

AcademicHub 汇聚顶级期刊与预印本平台的实时文献。定制您的专属科研雷达,利用大语言模型自动生成交叉领域文献分析简报。

01.
arXiv (CS.LG) 2026-06-11

Conformal Bayes under Label Shift: Post-Hoc Calibration vs. In-Training Adaptation

arXiv:2606.11865v1 Announce Type: cross Abstract: Conformal Bayes combines Bayesian posterior predictives with conformal calibration to produce prediction sets that are both statistically valid and geometrically efficient. We study conformal Bayes under label shift from a unified perspective, identifying two complementary approaches that restore nominal target-domain coverage through importance-weighted conformal calibration but operate through independent mechanisms. Post-hoc calibration tilts the posterior predictive toward the target domain and corrects the conformal threshold via an importance-weighted quantile, leaving the parameter posterior unchanged. In-training adaptation tilts the parameter posterior itself to the target domain, producing a corrected predictive whose highest predictive density region serves as the highest predictive density (HPD) based prediction set under the fitted target predictive; efficiency is model-dependent and does not imply finite-sample conditional optimality. Two controlled experiments show that in an unbiased training regime both strategies achieve valid coverage equally, while in a lead-optimization regime in-training adaptation acts as a debiasing operator, reducing interval width at unchanged coverage.

02.
arXiv (CS.CL) 2026-06-19

Analyzing Error Propagation in Korean Spoken QA with ASR-LLM Cascades

We analyze how automatic speech recognition (ASR) errors propagate through ASR-LLM cascades in Korean spoken question answering (SQA), focusing on downstream semantic failures that conventional ASR metrics cannot fully capture. Our analysis shows that the relative downstream degradation caused by ASR errors is consistent across LLMs with different absolute performance, suggesting that cascade degradation largely tracks ASR-stage information loss. We further identify single-character Korean ASR errors as a Korean-specific loss channel, where even a minimal transcription difference can change the intended question and degrade downstream QA performance. Finally, an auxiliary comparison shows that a large audio language model outperforms an ASR-LLM cascade with an approximately matched language backbone in noisy Korean SQA, indicating the potential of direct audio input to mitigate transcript-induced information loss.

03.
arXiv (CS.AI) 2026-06-19

CRAX: Fast Safe Reinforcement Learning Benchmarking

arXiv:2606.20376v1 Announce Type: cross Abstract: Safety is a core concern for deploying reinforcement learning (RL) agents in real-world domains such as robotics and autonomous driving. While benchmarks have been central to progress in RL, existing safety benchmarks with high-fidelity 3D physics remain computationally slow, limiting large-scale experimentation and rapid prototyping. To address this gap, we propose CRAX (Constrained RL Accelerated with JAX). Built on top of the MuJoCo XLA (MJX) physics engine with realistic 3D dynamics, CRAX leverages vectorized operations and hardware acceleration, yielding up to ~100x speedups over comparable CPU-based safety benchmarks. The benchmark features six environment suites and three agent-specific tasks, each spanning three difficulty levels. Evaluating six popular safe RL methods shows that no single approach dominates across all tasks, and reveals the trade-offs between performance and safety. We find that curriculum learning across difficulty levels and safety transfer can improve performance over direct training in harder settings.

04.
medRxiv (Medicine) 2026-06-11

Long-term exposure to PM2.5 components and lipid profiles in WTC Health Program general responders

Fine particulate matter (PM2.5) was found to be associated with elevated blood lipids, but fewer studies have examined the associations with specific constituents of PM2.5. We studied the associations between exposure to annual PM2.5 and its 14 constituents, and repeated blood lipid measurements among general responders enrolled in the World Trade Center Health Program between 2003 and 2019 (n = 44,876). We used generalized additive mixed effect models to investigate the single-pollutant associations with repeated measures of blood total cholesterol (TC), high and low-density lipoprotein (HDL-C and LDL-C) levels. We then used linear generalized weighted quantile sum regression with a random intercept for participant ID to account for the clustering of repeated measures and evaluate the combined associations with the component mixture. A decile increase in the mixture of 14 PM2.5 chemical components was associated with 0.375 mg/dL increase in TC levels (95% confidence Interval (CI): 0.174-0.577) and 0.302 mg/dL increase in LDL-C (95% CI: 0.063, 0.540). Lead, organic carbon, and iron were major drivers of both associations. Component-specific models also show higher TC and LDL levels associated with interquartile range increases in organic carbon (0.472, 95% CI [0.027, 0.918] and 0.648 95% CI [0.136, 1.160]) and iron exposure (1.081, 95% CI [0.630, 1.532] and 0.748, 95% CI [0.318, 1.178]). In conclusion, we found PM2.5 exposure to be associated with elevated lipid levels. The associations differed by PM2.5 composition, highlighting organic carbon, lead, and iron and major drivers. These findings are highly significant for a population exposed to extreme air pollution event and susceptible to lipid alterations that might trigger cardiovascular events.

05.
arXiv (CS.CV) 2026-06-17

Million-scale multimodal pollen microscopy with expert-guided foundation models

Automated pollen identification from microscopy remains a bottleneck in aerobiology, palaeoecology and biodiversity monitoring, because scalable systems must generalise across specimen preparation, scanner settings and geographic origins while retaining palynological interpretability. To address this gap, we present a million-scale multimodal pollen microscopy resource, Pollen AI Atlas, assembled from pure-species whole-slide bright-field images spanning four geographic origins, four scanner settings and 46 taxon labels across 31 botanical families. Seeded by one manually selected exemplar per source slide, token-level mining and filtering produced 1,511,390 released grain detections with 99.6\% proposal precision in expert-curated test regions. Each detection was paired with machine-generated grain-level morphological captions from five open-weight vision-language models, guided by expert-verified palynological anchors, yielding structured descriptions of aperture systems, wall ornamentation, shape and size. Among the evaluated models, Gemma4 provided the most controlled primary caption set, combining tight length control, no leakage and the strongest text-retrieval performance. Baseline benchmarks with frozen visual features reached 88.16\% top-1 accuracy, while cross-regional retrieval showed that caption-derived text embeddings remained robust when image similarity degraded (mAP@20 0.811 versus 0.262). Released data, annotations, captions, splits, code, and weights provide a benchmark for pollen recognition, cross-regional domain adaptation and domain-specific multimodal microscopy learning.

06.
arXiv (CS.CL) 2026-06-17

Position: Coding Benchmarks Are Misaligned with Agentic Software Engineering

Coding agents have become a major mode of software engineering, but the benchmarks we use to compare them were designed in a pre-agent era: they collapse model, harness, and environment into a single end-to-end score, typically computed against one reference solution, with no component-level signal for iteration. We argue that current coding benchmarks are misaligned with agentic software engineering. A coding agent in practice is not a model: it is a system harness – a composite of models, harnesses, contexts, environments, and feedback signals, any one of which can move the benchmark score by margins comparable to those between adjacent model generations. We discuss three symptoms: (i) benchmark scores conflate the model with the rest of the harness; (ii) grading against a single reference solution penalises equally valid alternatives; and (iii) the absence of signal at the level of individual harness components makes the end-to-end system score difficult to iterate on.

07.
arXiv (CS.CV) 2026-06-16

Token-Level Entropy Reveals Demographic Disparities in Language Models

We ask whether demographic identity, signaled by a name alone, systematically reshapes the generative distribution of a language model. Measuring full-vocabulary Shannon entropy at temperature zero across six open-weight base models and 5,760 implicit sentence-completion prompts (e.g., "Tanisha walked into the office on a Monday morning and"), we find that Black-associated names produce higher first-token entropy than White-associated names across all six architectures - opposite to the output-level homogeneity bias documented under explicit demographic prompting (Lee et al., 2024) - and Black-associated names always produce greater entropy above identity-neutral baselines than White-associated names ($\Delta\Delta > 0$ in all six models). Women-associated names co-occur with lower first-token entropy (DL-pooled $\hat\beta = -0.041, p = .019$) and more homogeneous outputs ($\hat\alpha = +0.024, p < .001$) than men-associated names - a pattern convergent with homogeneity bias; race and gender effects are additive. Instruction tuning does not attenuate the race gap (matched-format DL-pooled $\hat{\beta}=+0.153$). Running the same templates with explicit group labels instead of names yields null race effects in 10 of 12 models where implicit probing is significant - establishing that probing methodology is a primary determinant of which distributional structure is recovered.

08.
arXiv (CS.AI) 2026-06-11

EvalStop: Using World Feedback to Detect and Correct Reward Overoptimization in Multi-Tenant RLHF Platforms

arXiv:2606.04145v2 Announce Type: replace-cross Abstract: Cloud LLM fine-tuning platforms increasingly serve RLHF workloads, where a learned reward model is optimized as a proxy for human quality. As Gao et al. (2023) showed, this proxy diverges from world feedback (downstream eval metrics) under sustained optimization pressure, a phenomenon known as reward overoptimization. Existing platform schedulers ignore this divergence: non-clairvoyant schedulers optimize JCT without any quality signal, SLAQ-style quality-aware schedulers use training loss (a weaker proxy that drops monotonically through hacking), and classical per-job early stopping requires human monitoring and does not free shared GPUs. We propose EvalStop, a composable scheduling primitive that terminates jobs on k consecutive eval-score declines, releases GPUs, preserves the best checkpoint, and delegates to any base scheduler. We frame scheduler-level early stopping as a detection problem and evaluate it in a discrete-event simulator whose RLHF workload mixes reward-hacking and structurally healthy runs, with ground-truth labels hidden from schedulers. On RLHF-heavy workloads (80% RLHF, 64 GPUs), EvalStop achieves precision 98% / recall 99% / FPR 1.5% while improving JCT by 9% and cutting wasted compute by 22% over SRTF-Est (p

09.
arXiv (CS.AI) 2026-06-17

Catastrophic Forgetting is Low-Rank: A Function-Space Theory for Continual Adaptation

arXiv:2606.18024v1 Announce Type: cross Abstract: Catastrophic forgetting in continual adaptation is usually studied through parameter drift, replay, or distillation, but these views do not identify which output-space directions are vulnerable. We give a function-space account in the NTK regime: new-task training induces old-task prediction drift through the cross-task kernel, yielding a closed-form predictor for the forgetting vector before any new-task gradient step. In frozen-backbone linear-head PEFT-CL, where the model is linear in the trainable parameters, the predictor is exact up to numerical precision; for nonlinear adapters/full fine-tuning, it is a local NTK approximation. The same expression reveals that forgetting concentrates in a small number of old-task NTK eigenmodes and under frozen linear heads gives a Kronecker scaling rule for the vulnerable rank. These results clarify the relation to prior NTK-overlap theory, explain why parameter-space regularizers can miss output-space interference, and motivate a targeted spectral regularizer.

10.
medRxiv (Medicine) 2026-06-17

Proteomics Uncovers Cryptic JPH2 Loss in Paediatric Dilated Cardiomyopathy

Despite recent advances in next-generation sequencing, genetic diagnostic rates for dilated cardiomyopathy (DCM) remain low. Among paediatric DCM, causes are often heritable, with a greater frequency of de novo, recessive and syndromic causes of disease. Novel diagnostic methods are therefore required to solve monogenic cases. To assess the value of proteomics as a diagnostic tool for paediatric DCM, we obtained left ventricle myocardial samples from paediatric patients undergoing heart transplantation at the Royal Children's Hospital, Melbourne. We performed genome sequencing and proteomics and leveraged this multi-omics dataset to uncover the molecular cause of disease in a gene elusive proband. The proband carried a heterozygous JPH2 frameshift variant identified on clinical exome sequencing. However, proteomic analysis showed a pronounced downregulation of JPH2, suggestive of biallelic loss-of-function. Closer inspection of the genomic data revealed a large inversion (~8.34 Mb) with a breakpoint falling within intron 5 of JPH2 that displaces the 3'UTR from the coding transcript. The two variants were confirmed to be in trans using long read DNA sequencing, consistent with a diagnosis of JPH2 autosomal recessive DCM. Finally, we applied RNA sequencing with total RNA library preparation to show that transcripts containing a 3'UTR were reduced to ~10% relative to controls. As a proof-of-principle, we present the first reported use of proteomics from explanted cardiac tissue to provide a genetic diagnosis. Our methodology has broad relevance to patients with genetically unsolved Mendelian diseases, who might undergo organ transplantation as part of clinical management.

11.
arXiv (CS.LG) 2026-06-11

Holding the FP8 Quality Ceiling at 8-Bit Weights and Activations: INT8 and GGUF Post-Training Quantization of Ideogram 4.0 for Consumer GPUs

arXiv:2606.12280v1 Announce Type: new Abstract: Post-training quantization lets large text-to-image diffusion transformers run on consumer GPUs, yet the hardware-specific trade-offs are seldom measured directly. We quantize Ideogram 4.0 - a 9.3B flow-matching diffusion transformer (DiT), shipped as two separate-weight copies of a single-stream 34-layer backbone for classifier-free guidance and conditioned by a Qwen3-VL-8B encoder - for Ampere RTX 3090 GPUs, which lack FP8 tensor cores. Our INT8 W8A8 recipe (per-channel weights, per-token dynamic activations, SmoothQuant, and mixed-precision protection of a small high-fragility layer set) holds the FP8 quality ceiling: on a 200-prompt benchmark the paired same-seed bootstrap CI for INT8-FP8 includes zero on both Pick and CLIP, while INT8 improves on NF4 by $+1.9$ CLIP (95% CI $[+1.21,+2.64]$, excluding zero). A per-category OCR analysis, to our knowledge unreported for this model class, confirms text legibility is preserved, and an ablation isolates protection of the FFN down-projections as the dominant quality lever. Our GGUF Q4_K quantization beats NF4 at equal on-disk size and is the Pareto winner on the quality-memory frontier, with paired confidence intervals excluding zero (Q8_0 is quality neutral). Finally, we characterize where 8-bit quantization helps and where it does not: INT8's weights match FP8's footprint rather than shrink it, so a speed gain on Ampere awaits a fused INT8 kernel.

12.
arXiv (CS.CL) 2026-06-12

A Context-Aware Dataset for Stance Detection in Bioethical Controversies on Reddit

Bioethical debates increasingly unfold on social media, yet stance detection research lacks large-scale, domain-specific resources for modeling such context-dependent discourse. We present BioStance, a context-aware dataset of 39,600 annotated Post-Comment pairs from Reddit bioethical discussions. BioStance covers six controversial targets across three dimensions of bioethical controversy: fundamental value conflicts, individual liberty versus collective responsibility, and technological uncertainty. Each instance preserves hierarchical conversational context and is labeled by three independent annotators using a three-class stance scheme: Favor, Against, and None. The annotations achieve a mean Krippendorff's $\alpha$ of 0.82, indicating substantial reliability. By combining thematic diversity, conversational structure, and high-quality human annotation, BioStance supports research on context-aware stance detection, argument mining, and computational analysis of bioethical discourse.

13.
arXiv (CS.CV) 2026-06-17

Beware of Aliases – Signal Preservation is Crucial for Robust Image Restoration

Image restoration networks are usually comprised of an encoder and a decoder, responsible for aggregating image content from noisy, distorted data and to restore clean, undistorted images, respectively. Data aggregation as well as high-resolution image generation both usually come at the risk of involving aliases, i.e.~standard architectures put their ability to reconstruct the model input in jeopardy to reach high PSNR values on validation data. The price to be paid is low model robustness. In this work, we show that simply providing alias-free paths in state-of-the-art reconstruction transformers supports improved model robustness at low costs on the restoration performance. We do so by proposing BOA-Restormer, a transformer-based image restoration model that executes downsampling and upsampling operations partly in the frequency domain to ensure alias-free paths along the entire model while potentially preserving all relevant high-frequency information.

14.
arXiv (math.PR) 2026-06-16

Logarithmic Large Deviations for Heavy-Tailed Sums

arXiv:2606.16487v1 Announce Type: new Abstract: We establish logarithmic large-deviation bounds for sums of independent nonnegative random variables with regularly varying tails. The normalization is chosen at the extreme-value scale and the speed is $\log n$. In contrast with Cramér's theorem, the resulting rate function is determined only by the tail index. The proof transfers a maximum large-deviation principle to sums in the one-big-jump region.

15.
arXiv (CS.CV) 2026-06-11

PIGEON: VLM-Driven Object Navigation via Points of Interest Selection

Object navigation in unseen indoor environments requires agents to perform semantic search under partial observability. Vision-language models (VLMs) provide strong semantic-spatial priors for this task, but how to interface them with robot navigation remains challenging: dense VLM inference is expensive, while abstracting environments into symbolic memories often separates high-level reasoning from the raw visual evidence that supports it. We propose we propose PIGEON (Point of Interest Guided Exploration for Object Navigation), a VLM-driven framework that formulates object navigation as raw-observation-grounded sparse decision problem. PIGEON introduces Points of Interest (PoIs) as sparse visual decision units that couple geometrically executable waypoints with raw egocentric observations. Rather than using VLMs as dense controllers or restricting them to frontier ranking, PIGEON enables VLMs to select among task-critical PoIs, including exploration frontiers, suspected target objects, traversable stairs, and floor-level summaries, while low-level planners execute continuous motion between them. This PoI interface further makes high-level navigation decisions verifiable, allowing us to develop an RLVR pipeline that improves local VLMs without manual Chain-of-Thought annotations. Extensive experiments on Habitat ObjectNav benchmarks show that PIGEON achieves state-of-the-art zero-shot performance, scales consistently with foundation model capacity, and transfers to Active Embodied Question Answering with only prompt modifications. Real-world deployments on physical robots further demonstrate its robustness and efficiency.

16.
arXiv (quant-ph) 2026-06-12

Matrix phase-space representations for quantum symmetries

arXiv:2606.12769v1 Announce Type: new Abstract: We introduce a general phase-space representation that includes global quantum symmetries in the basis expansion. This method, called matrix phase-space, projects the basis onto a reduced Hilbert space, which can greatly reduce sampling errors of many-body quantum simulations and unifies several previous phase-space methods. The purpose of this paper is to provide detailed proofs of basic theorems and operator identities. We also treat several different types of symmetries. To illustrate the benefits of matrix phase-space methods, we give a detailed derivation of a recent application to the topical problem of verifying the outputs of Gaussian boson sampling (GBS) quantum computers with photon number resolving detectors. This has exponential complexity, and using parity symmetry reduces sampling errors by very large factors relative to earlier methods.

17.
arXiv (CS.CV) 2026-06-18

NeuMesh++: Towards Versatile and Efficient Volumetric Editing with Disentangled Neural Mesh-based Implicit Field

Recently neural implicit rendering techniques have evolved rapidly and demonstrated significant advantages in novel view synthesis and 3D scene reconstruction. However, existing neural rendering methods for editing purposes offer limited functionalities, e.g., rigid transformation and category-specific editing. In this paper, we present a novel mesh-based representation by encoding the neural radiance field with disentangled geometry, texture, and semantic codes on mesh vertices, which empowers a set of efficient and comprehensive editing functionalities, including mesh-guided geometry editing, designated texture editing with texture swapping, filling and painting operations, and semantic-guided editing. To this end, we develop several techniques including a novel local space parameterization to enhance rendering quality and training stability, a learnable modification color on vertex to improve the fidelity of texture editing, a spatial-aware optimization strategy to realize precise texture editing, and a semantic-aided region selection to ease the laborious annotation of implicit field editing. Extensive experiments and editing examples on both real and synthetic datasets demonstrate the superiority of our method on representation quality and editing ability. Project page: https://zju3dv.github.io/neumeshplusplus/

19.
arXiv (quant-ph) 2026-06-17

Split-Head Quantum Generative Adversarial Network for Crystalline Material Discovery

arXiv:2606.17852v1 Announce Type: new Abstract: The discovery of novel crystalline materials is a critical challenge in computational materials science, often limited by the spatial representation limitations and mode collapse typical of classical generative models. Traditionally, developing Quantum GANs for continuous 3D space is hindered by the limited capacity of near-term hardware. To overcome this, we adapt a physics-informed "split-head" architecture right from the quantum trunk to explicitly decouple macroscopic lattice bounds from microscopic atomic coordinates, significantly maximizing resource efficiency. This study disentangles the contributions of quantum circuits from these architectural priors by evaluating a Split-Head Quantum Generative Adversarial Network against an architecture-matched classical ablation model. Evaluated on the highly constrained Mg-Mn-O system, the results reveal a highly nuanced performance dichotomy between the advanced models. The architecture-matched classical ablation model demonstrated superior thermodynamic precision. Conversely, the integration of quantum circuits in the SH-QGAN drove unparalleled structural breadth and latent space exploration, more than doubling the ablation's geometric validity and successfully generating novel, metastable candidates converging on the Mg2MnO4 stoichiometry. These findings clarify that while architectural separation of cell and atom generation drives strict thermodynamic precision, quantum feature mapping independently provides the spatial diversity necessary to overcome mode collapse. Both mechanisms offer distinct, complementary enhancements for the generative discovery of advanced materials.

20.
arXiv (CS.LG) 2026-06-15

Can Machine Learning Forecast Rice Yields in Data-Constrained Settings? Satellite Climate Data, National Crop Statistics, and Lessons from Sierra Leone

arXiv:2606.13959v1 Announce Type: new Abstract: Sierra Leone's agriculture operates with almost no data-driven decision support, and no published machine learning study has examined the country's crop yields. We ask whether rice yield can be forecast from data Sierra Leone currently has. Using 25 years of FAOSTAT production data (2000-2024) for nine major crops, we train XGBoost, Gradient Boosting, and Random Forest under a strict anti-leakage protocol with expanding-window walk-forward evaluation across seven held-out years, benchmarked against naive persistence. No model trained on crop statistics alone outperforms persistence. Augmenting with free satellite climate data (CHIRPS rainfall, NASA POWER temperature) reverses this result: a climate-only XGBoost reduces forecast error by one third (RMSE 284 vs 428 kg/ha), a gain that holds for a linear model and is robust to excluding the anomalous 2018 season. Early-season (May-June) rainfall is the dominant predictor, implying seasonal yield risk is observable months before harvest. No model anticipated the 2018 collapse, whose origins were institutional rather than climatic. We translate the findings into policy recommendations for Sierra Leone's Feed Salone Strategy, with a fully open-source pipeline.

21.
arXiv (CS.LG) 2026-06-16

ROVE: Unlocking Human Interventions for Humanoid Manipulation via Reinforcement Learning

arXiv:2606.17011v1 Announce Type: cross Abstract: Human interventions provide crucial corrective signals for post-training Vision-Language-Action (VLA) models. However, enabling seamless humanoid interventions is a formidable systems challenge due to complex whole-body kinematics and dexterous-hand control. Consequently, the collected intervention trajectories are often suboptimal, and methods that rely on human interventions as expert supervision can absorb hesitant, inefficient, or even erroneous behaviors. To address both the system and algorithmic challenges, we propose ROVE, a reinforcement learning framework for humanoid VLA post-training with imperfect human interventions. First, ROVE introduces a human-in-the-loop pipeline capable of collecting deployment and intervention data for humanoid manipulation. Second, it utilizes Optimistic Value Estimation (OVE) to prioritize high-value behaviors from mixed-quality trajectories. To further robustify value estimation, we incorporate cross-embodiment human experience videos to provide rich supervision for long-tailed failure and recovery modes. The resulting critic yields informative advantage signals, steering the VLA actor to focus on high-value behaviors rather than indiscriminately imitating all actions. On challenging real-world contact-rich and fine-grained humanoid manipulation tasks, ROVE outperforms experience-learning baselines and consistently improves across multiple rollout-intervention iterations.

23.
arXiv (CS.CV) 2026-06-18

A Unified Framework for Efficient Remote Sensing Visual Question Answering: Adapting Dual, Hybrid, and Encoder-Decoder Architectures

Visual Question Answering (VQA) in the Remote Sensing (RS) domain presents unique challenges due to the high resolution, multi scale object distribution, and semantic complexity of aerial imagery. While general domain Foundation Models have achieved remarkable success, their direct application to RSVQA is hindered by massive domain shifts and the computationally prohibitive nature of full fine tuning. This study presents a comparative analysis of RS Adapter, a Parameter Efficient Fine Tuning (PEFT) strategy, applied across three distinct Vision Language Model (VLM) architectures: the Dual Encoder CLIP, the Encoder Decoder BLIP, and the Hybrid FLAVA. We introduce a unified architectural surgery pipeline that injects lightweight bottleneck adapters into the attention and MLP layers of frozen backbones, enabling rapid adaptation with less than 5 percent of trainable parameters. Experimental results on the high resolution RSVQA x dataset demonstrate that while all adapted models achieve convergence, the Hybrid FLAVA architecture offers a superior balance of multimodal reasoning and retrieval capabilities compared to its unimodal counterparts. Our findings establish a new baseline for resource efficient VQA in disaster assessment and urban monitoring.

24.
arXiv (CS.AI) 2026-06-16

Honeypot Protocol

作者:

arXiv:2604.13301v1 Announce Type: cross Abstract: Trusted monitoring, the standard defense in AI control, is vulnerable to adaptive attacks, collusion, and strategic attack selection. All of these exploit the fact that monitoring is passive: it observes model behavior but never probes whether the model would behave differently under different perceived conditions. We introduce the honeypot protocol, which tests for context-dependent behavior by varying only the system prompt across three conditions (evaluation, synthetic deployment, explicit no-monitoring) while holding the task, environment, and scoring identical. We evaluate Claude Opus 4.6 in BashArena across all three conditions in both honest and attack modes. The model achieved 100% main task success and triggered zero side tasks uniformly across conditions, providing a baseline for future comparisons with stronger attack policies and additional models.

25.
arXiv (CS.CL) 2026-06-16

DRA-GRPO: Your GRPO Needs to Know Diverse Reasoning Paths for Mathematical Reasoning

Post-training LLMs with Reinforcement Learning, specifically Group Relative Policy Optimization (GRPO), has emerged as a paradigm for enhancing mathematical reasoning. However, standard GRPO relies on scalar correctness rewards that are often non-injective with respect to semantic content: distinct reasoning paths receive identical rewards. This leads to a Diversity-Quality Inconsistency, where the policy collapses into a narrow set of dominant modes while ignoring equally valid but structurally novel strategies. To bridge this gap, we propose Diversity-aware Reward Adjustment (DRA), a theoretically grounded framework that calibrates the reward signal using the semantic density of sampled groups. By leveraging Submodular Mutual Information (SMI), DRA implements an Inverse Propensity Scoring (IPS) mechanism that effectively de-biases the gradient estimation. This creates a repulsive force against redundancy, driving the policy to achieve better coverage of the high-reward landscape. Our method is plug-and-play and integrates seamlessly with GRPO variants. Empirical evaluations on five math benchmarks demonstrate that DRA-GRPO consistently outperforms strong baselines, achieving an average accuracy of 58.2% on DeepSeek-R1-Distill-Qwen-1.5B with only 7,000 training samples and $55 cost, highlighting the critical role of diversity calibration in data-efficient alignment. The code is available at https://github.com/xiwenc1/DRA-GRPO.