Agents' Last Exam

Recent AI systems have achieved strong results on a wide range of benchmarks, yet these gains have not translated into economically meaningful deployment across many professional domains. We argue that this gap is largely an evaluation problem: widely used benchmarks lack sustained performance measurement on real and economically valuable workflows. This paper introduces Agents' Last Exam (ALE), a benchmark designed to evaluate AI agents on long-horizon, economically valuable, real-world tasks with verifiable outcomes. Developed in collaboration with 250+ industry experts, ALE covers non-physical industries defined with reference to O*NET / SOC 2018 (the U.S. federal occupational taxonomy). It is organized around a task taxonomy with 55 subfields grouped into 13 industry clusters covering 1K+ tasks. Current results show that the hardest tier remains far from saturated: across mainstream harness and backbone configurations, the average full pass rate is 2.6%. ALE is designed as a living benchmark: its task pool grows continuously as new workflows and industries are onboarded. More broadly, ALE is intended not merely as another leaderboard, but as an instrument for closing the gap between benchmark success and GDP-relevant impact.

PolySpeech-100: A Large-Scale Benchmark for Speech Understanding Across 100+ Languages and Dialects

While End-to-End (E2E) Speech-Large Language Models (Speech-LLMs) are rapidly evolving, their evaluation methodologies remain limited to the era of simple transcription. Existing benchmarks suffer from three critical limitations: a pronounced bias towards high-resource languages, a focus on low-level recognition (ASR) rather than semantic reasoning, and a neglect of regional dialects. To bridge this gap, we introduce PolySpeech-100, a massive-scale benchmark designed to assess `native-level' speech comprehension across 110 linguistic variants. We employ a novel hybrid construction pipeline that augments gold-standard human recordings with instruction-driven synthetic speech, allowing us to cover 19 distinct Chinese dialects and over 80 low-resource languages. Extensive evaluation of 22 state-of-the-art models (including Gemini-3, GPT-Audio, and Qwen2.5-Omni) yields pivotal insights. First, we demonstrate that open-source E2E models outperform Cascade (ASR+LLM) systems on heavy dialects, proving that direct audio processing preserves critical paralinguistic cues and prosodic features (e.g., intonation, stress) that are often lost in standard transcription. Second, we reveal a significant performance gap: while commercial models maintain robustness, open-source models suffer catastrophic degradation on low-resource languages. Finally, counter-intuitively, we observe that under standard zero-shot settings, Chain-of-Thought prompting frequently degrades speech understanding performance for most evaluated models, revealing a potential modality alignment gap in current architectures. PolySpeech-100 establishes a rigorous standard for the next generation of inclusive, omni-capable Speech-LLMs. The data, demo, and code are publicly available at https://github.com/YoungSeng/PolySpeech-100.

Fisher-Preserving Guidance: Training-Free Manifold Constraints for Safe Diffusion Control

Diffusion models are effective for waypoint prediction in visual navigation, but standard sampling and test time guidance can produce unreliable or inefficient trajectories when updates drift off the training manifold. We propose Fisher Preserving Guidance with Outer Product Span Projection, a training-free inference method that avoids large Fisher drift associated with off-distribution actions while optimizing a task objective. Our method computes the Fisher-preserving update via a low-rank Jacobian factorization, requiring only a single backward pass per step and enabling real-time use. We further introduce Truncated Fisher Denoising Sensitivity as an uncertainty signal and use it for robust multi-sample action blending. Experiments on toy and realistic navigation benchmarks, including Maze2D with TSDF-based guidance, PushT with official Diffusion Policy weights, and visual navigation in simulation and on real robots, demonstrate consistent improvements in performance over strong diffusion-policy baselines without additional training.

MUSE: Benchmarking Manufacturable, Functional, and Assemblable Text-to-CAD Generation

Large language models (LLMs) have recently advanced text-driven 3D generation, yet Text-to-CAD remains far from supporting industrial product design. Existing benchmarks focus primarily on generating single-part CAD models and evaluate them using geometric similarity metrics that fail to capture functionality, manufacturability, and assemblability. To address this gap, we introduce MUSE, a Text-to-CAD benchmark focused on complex, editable boundary representation (B-Rep) assemblies. MUSE pairs practical design instances with structured Design Specifications and evaluates generated models through a three-stage protocol: code check, geometric check, and design-intent alignment. The final stage uses design-specific rubrics to assess functionality, manufacturability, and assemblability, moving beyond shape matching toward practical design quality. To enable scalable evaluation, we use a rubric-based visual language model (VLM) judge and validate its reliability through human annotation. Experiments on closed-source and open-source LLMs reveal a clear failure cascade from executable code to valid geometry and finally to engineering-ready design, with even the strongest models achieving limited success on fine-grained engineering criteria. Together, MUSE provides a realistic benchmark and evaluation framework for advancing Text-to-CAD from geometric generation toward true engineering design. Our project website, including the leaderboard, dataset, and code, is available at https://dong7313.github.io/muse-benchmark/.

Random Walk on Point Clouds for Feature Detection

The points on the point clouds that can entirely outline the shape of the model are of critical importance, as they serve as the foundation for numerous point cloud processing tasks and are widely utilized in computer graphics and computer-aided design. This study introduces a novel method, RWoDSN, for extracting such feature points, incorporating considerations of sharp-to-smooth transitions, large-to-small scales, and textural-to-detailed features. We approach feature extraction as a two-stage context-dependent analysis problem. In the first stage, we propose a novel neighborhood descriptor, termed the Disk Sampling Neighborhood (DSN), which, unlike traditional spatially and geometrically invariant approaches, preserves a matrix structure while maintaining normal neighborhood relationships. In the second stage, a random walk is performed on the DSN (RWoDSN), yielding a graph-based DSN that simultaneously accounts for the spatial distribution, topological properties, and geometric characteristics of the local surface surrounding each point. This enables the effective extraction of feature points. Experimental results demonstrate that the proposed RWoDSN method achieves a recall of 0.769-22% higher than the current state-of-the-art-alongside a precision of 0.784. Furthermore, it significantly outperforms several traditional and deep-learning techniques across eight evaluation metrics.

WebWorld: A Large-Scale World Model for Web Agent Training

Web agents require massive trajectories to generalize, yet real-world training is constrained by network latency, rate limits, and safety risks. We introduce \textbf{WebWorld} series, the first open-web simulator trained at scale. While existing simulators are restricted to closed environments with thousands of trajectories, WebWorld leverages a scalable data pipeline to train on 1M+ open-web interactions, supporting reasoning, multi-format data, and long-horizon simulations of 30+ steps. For intrinsic evaluation, we introduce WebWorld-Bench with dual metrics spanning nine dimensions, where WebWorld achieves simulation performance comparable to Gemini-3-Pro. For extrinsic evaluation, Qwen3-14B trained on WebWorld-synthesized trajectories improves by +9.2\% on WebArena, reaching performance comparable to GPT-4o. WebWorld enables effective inference-time search, outperforming GPT-5 as a world model. Beyond web simulation, WebWorld exhibits cross-domain generalization to code, GUI, and game environments, providing a replicable recipe for world model construction.

ERNIE 5.0 Technical Report

In this report, we introduce ERNIE 5.0, a natively autoregressive foundation model desinged for unified multimodal understanding and generation across text, image, video, and audio. All modalities are trained from scratch under a unified next-group-of-tokens prediction objective, based on an ultra-sparse mixture-of-experts (MoE) architecture with modality-agnostic expert routing. To address practical challenges in large-scale deployment under diverse resource constraints, ERNIE 5.0 adopts a novel elastic training paradigm. Within a single pre-training run, the model learns a family of sub-models with varying depths, expert capacities, and routing sparsity, enabling flexible trade-offs among performance, model size, and inference latency in memory- or time-constrained scenarios. Moreover, we systematically address the challenges of scaling reinforcement learning to unified foundation models, thereby guaranteeing efficient and stable post-training under ultra-sparse MoE architectures and diverse multimodal settings. Extensive experiments demonstrate that ERNIE 5.0 achieves strong and balanced performance across multiple modalities. To the best of our knowledge, among publicly disclosed models, ERNIE 5.0 represents the first production-scale realization of a trillion-parameter unified autoregressive model that supports both multimodal understanding and generation. To facilitate further research, we present detailed visualizations of modality-agnostic expert routing in the unified model, alongside comprehensive empirical analysis of elastic training, aiming to offer profound insights to the community.

Not All Negative Samples Are Equal: LLMs Learn Better from Plausible Reasoning

Learning from negative samples holds great promise for improving Large Language Model (LLM) reasoning capability, yet existing methods treat all incorrect responses as equally informative, overlooking the crucial role of sample quality. To address this, we propose Plausible Negative Samples (PNS), a method that synthesizes high-quality negative samples exhibiting expected format and structural coherence while ultimately yielding incorrect answers. PNS trains a dedicated model via reverse reinforcement learning (RL) guided by a composite reward combining format compliance, accuracy inversion, reward model assessment, and chain-of-thought evaluation, generating responses nearly indistinguishable from correct solutions. We further validate PNS as a plug-and-play data source for preference optimization across three backbone models on seven mathematical reasoning benchmarks. Results demonstrate that PNS consistently outperforms other negative sample synthesis methods, achieving an average improvement of 2.03% over RL-trained models.

A Hitchhiker's Guide to Poisson Gradient Estimation

Poisson-distributed latent variable models are widely used in computational neuroscience, but differentiating through discrete stochastic samples remains challenging. Two approaches address this: Exponential Arrival Time (EAT) simulation and Gumbel-SoftMax (GSM) relaxation. We provide the first systematic comparison of these methods, along with practical guidance for practitioners. Our main technical contribution is a modification to the EAT method that theoretically guarantees an unbiased first moment (exactly matching the firing rate), and reduces second-moment bias. We evaluate these methods on their distributional fidelity, gradient quality, and performance on two tasks: (1) variational autoencoders with Poisson latents, and (2) partially observable generalized linear models, where latent neural connectivity must be inferred from observed spike trains. Across all metrics, our modified EAT method exhibits better overall performance (often comparable to exact gradients), and substantially higher robustness to hyperparameter choices. Together, our results clarify the trade-offs between these methods and offer concrete recommendations for practitioners working with Poisson latent variable models.

Human-in-the-Loop Failure Recovery with Adaptive Task Allocation

Since the recent Covid-19 pandemic, mobile manipulators and humanoid assistive robots with higher levels of autonomy have increasingly been adopted for patient care and living assistance. Despite advancements in autonomy, these robots often struggle to perform reliably in dynamic and unstructured environments and require human intervention to recover from failures. Effective human-robot collaboration is essential to enable robots to receive assistance from the most competent operator, in order to reduce their workload and minimize disruptions in task execution. In this paper, we propose an adaptive method for allocating robotic failures to human operators (ARFA). Our proposed approach models the capabilities of human operators, and continuously updates these beliefs based on their actual performance for failure recovery. For every failure to be resolved, a reward function calculates expected outcomes based on operator capabilities and historical data, task urgency, and current workload distribution. The failure is then assigned to the operator with the highest expected reward. Our simulations and user studies show that ARFA outperforms random allocation, significantly reducing robot idle time, improving overall system performance, and leading to a more distributed workload among operators.