Search research

The synthesis of inductive loop invariants remains a critical bottleneck in automated program verification. While Large Language Models (LLMs) show promise in mitigating this issue, they often fail on complex programs, producing invariants that are invalid or computationally ineffective. Although fine-tuning is a natural strategy to address these limitations, obtaining high-quality training data remains an open challenge. We first formalize the properties required for a high-quality training invariant, and then present Wonda, a rigorous data curation pipeline that extracts such invariants from raw verifier output via AST-based normalization followed by LLM-driven semantic rewriting and augmentation with provable quality guarantees. Fine-tuning Small Language Models (SLMs) on Wonda-curated data yields consistent gains across the Qwen3, Llama-3.1, and Mistral families: the 4B and 8B Qwen3 models nearly double invariant correctness and double speedup rates, while Llama-3.1-8B triples both. On the challenging InvBench suite, the same 4B model outperforms an off-the-shelf model 20x its size and matches the end-to-end verification time of GPT-OSS-120B, while a 14B Qwen3 model matches that of the frontier model GPT-5.2, all without test-time compute overhead. Our code is publicly available on GitHub.

Post-training often improves task performance but can degrade confidence calibration, leaving post-trained language models (PoLMs) more overconfident than their corresponding pretrained language models (PLMs). Because task-specific labeled calibration data can be costly or unavailable, the corresponding pretrained PLM provides a natural label-free reference for post-hoc calibration. Prior agreement-gated PLM-referenced calibration fits a scalar temperature using only examples on which the PoLM and its PLM reference agree, excluding disagreement examples because direct alignment can drive the fitted temperature excessively high and induce under-confidence. We revisit this binary treatment. A controlled reintroduction diagnostic reveals a non-monotonic aggregate effect: admitting a moderate fraction of disagreement examples can improve calibration, whereas the benefit diminishes as unit-weight inclusion approaches the full disagreement set. We introduce SupportCal, a label-free post-hoc method that retains agreement examples at unit weight and assigns disagreement examples continuous weights based on the own-base PLM's relative support and corroboration from pretrained references selected from a size-compatible candidate pool. We further characterize when the resulting weighted objective admits a finite optimal temperature. Across MedMCQA and MathQA, SupportCal yields lower ECE than the agreement-only baseline for nearly all evaluated target-model configurations; supplementary TweetEval Sentiment results show the same pattern on a fixed-label classification task.

Vision-Language Pretrained Models (VLPMs) offer a scalable path to open-vocabulary chest radiology understanding, yet two aspects remain underexplored: how structured clinical semantics extracted from medical reports can reduce in-batch noise during contrastive learning, and how cross-modal fusion can be designed to produce more faithful spatial grounding without added complexity. We introduce AlphaRAD, addressing these opportunities through two contributions. First, we construct a large-scale structured medical concept space from medical reports parsed by a Large Language Model for training, thereby mitigating in-batch learning noise and removing heuristic pair matching in contrastive learning, and thus naturally positioning AlphaRAD as a medical concept discriminator trained via $α$-Corrected Binary Cross-Entropy. Second, we propose FLaS (Factorized Latent Supervision), an extremely simple yet effective cross-modal feature fusion module that factorizes VLPM representations into independent subspaces, using dedicated alignment supervision to enhance the expressiveness of spatial grounding without introducing additional model parameters. Through extensive empirical validation, AlphaRAD shows strong zero-shot generalization across diverse chest radiology tasks. Notably, it establishes state-of-the-art average performance across 16 classification benchmarks, while achieving individual state-of-the-art results via distinct gains on 7 grounding/phrase grounding and 3 segmentation datasets.

CorrNet is a strong baseline for continuous sign language recognition (CSLR) because it models inter-frame correlations inside the visual encoding stage. In this paper, we study two natural extensions of a reproduced CorrNet system: replacing the BiLSTM temporal head with a Transformer encoder, and injecting motion cues after temporal pooling. We find that the Transformer head does not outperform the BiLSTM baseline, even with a training strategy adjusted for the Transformer, and the two heads have almost the same computational and runtime cost. For the second extension, we design a lightweight module called MotionGate. In our experiments, MotionGate consistently collapses to an identity-like mapping: the gate loses motion selectivity, and the injected residual becomes a weak, non-selective perturbation of the pooled features. These results suggest that explicit motion injection after CorrNet's correlation-based encoding is largely redundant, and that natural-looking architectural extensions in CSLR should be tested carefully instead of being assumed to help.

While Large Language Model (LLM) agents demonstrate proficiency in static benchmarks, their deployment in real-world scenarios is hindered by the dynamic nature of user queries, tool sets, and interaction dynamics. To address this generalization gap, we formalize OpenAgent (Tool-Use Agent in Open-World), a problem setting characterized by distributional shifts across query, action, observation, and domain dimensions. To systematically diagnose its impact, we construct a controlled sandbox environment where we define fine-grained environmental shifts across a four-tier hierarchy, Perception, Interaction, Reasoning, and Internalization, and conduct a comprehensive series of experiments. Our analysis yields a series of key insights, demonstrating that agents trained via both Supervised Fine-Tuning(SFT) and Reinforcement Learning suffer from varying degrees of performance degradation when confronting open environmental shifts. Building on these insights, we propose Perturbation-Augmented Fine-Tuning, a disturbance-based intervention strategy for SFT that lays the foundation for enhancing agent robustness and utility in realistic environments. Our code will be released at: https://github. com/LAMDA-NeSy/OpenAgent.

Foreground segmentation is the critical first step of every computational pathology pipeline, yet existing methods rely on hand-tuned heuristics or supervised models that overfit to narrow stain and scanner distributions, failing silently on specialised stains such as Jones silver or Elastica van Gieson. We propose a coarse-to-fine approach that recasts foreground segmentation as a visual perception task and leverages general-purpose vision-language models (VLMs) as zero-annotation oracles. Our key insight is that tissue-versus-background discrimination is a natural-image recognition problem, not a histopathological one, so VLMs trained on internet-scale corpora generalise where domain-specific models cannot. We introduce Leica-75, a benchmark of 75 renal transplant whole-slide images spanning three stain families. On Leica-75, our method achieves the highest segmentation quality on out-of-distribution stains (Dice 0.858 +/- 0.027 on Jones, 0.853 +/- 0.041 on EVG) with 7x lower cross-stain variance than the best supervised baseline, while remaining competitive on in-distribution H&E. Few-shot prompting with automatically curated exemplars (Auto-context) rescues hard cases on Stress-32 (n=32), a curated stress-test subset (Dice 0.470 to 0.819 for the 2B model). VLM-based annotation review matches human expert consensus (kappa=0.989 for blur detection; mean precision/recall grading accuracy 0.708 vs. human 0.646 for segmentation mask review). The resulting pseudo-labels are used to distil lightweight student models that are as performant as the teacher model while running for a fraction of the cost. Our framework provides a principled, scalable solution to a persistent infrastructure bottleneck in digital pathology.

Deep learning models facilitate the discovery of molecules with tailored properties among billions of candidate compounds. However, the computational burden to develop and deploy state-of-the-art models continuously increases, limiting their scalability. Most large-scale models are unimodal in nature and overlook the potential to leverage complementary molecular data modalities. To address these shortcomings, this paper introduces the Graph-Language Alignment for Chemical Inference and Exploration using Representations (GLACIER) model, a student-teacher framework that integrates molecular graphs, SMILES strings, and physicochemical descriptors to learn rich molecular embeddings. Our framework consists of three stages: (1) we pretrain three student encoders on 100,000 drug-like molecules: a message-passing neural network for molecular graphs, a transformer-based encoder for SMILES strings, and a multilayer perceptron for physicochemical descriptors, (2) we fuse these student modalities using a novel Finsler geometry-aware module, and (3) distill complementary knowledge from large teacher models, including MiniMol and MolFormer, into a single lightweight model via contrastive learning. We demonstrate that GLACIER is a robust framework that delivers high predictive performance and computational efficiency in complex molecular property prediction tasks. Our code is publicly available at https://github.com/eemokey/glacier.

Tuning controllers for strongly coupled multi-input multi-output (MIMO) industrial processes is hard: decentralized classical auto-tuning ignores loop interaction, and local numerical optimization from natural initializations stalls in the resulting non-convex cost landscape. We ask whether on-premise open-source large language models (LLMs), which keep data on-site and need no plant model, can help. On a single-loop CSTR, classical relay-feedback tuning (IAE 0.106, near the 0.102 optimum) beats an LLM tuner (0.162): for simple loops the LLM adds nothing. The picture inverts on a strongly coupled quadruple-tank with conflicting set-points, scored by a penalized cost J = IAE + lambda*TV(u) that rewards tracking without chattering actuators. There, naive relay tuning (J ~ 28.6) and naive LLM tuning (29.7) are no better than open loop (22.7), and a local optimizer from balanced starts fails in 10/10 runs. A scaffolded open LLM instead reasons about the coupling, proposes the counter-intuitive asymmetric structure, and reaches J ~ 16.9 +/- 0.2 from any start; refining it with a classical optimizer attains the smooth global optimum (J ~ 12.0, 10/10 vs. 0/10), which even applies a non-obvious negative integral correction decentralized tuning cannot. A global optimizer (differential evolution) also reaches this optimum, so the LLM is not the only route; its advantage is sample efficiency and interpretability: a usable controller in 18 evaluations (where the global optimizer is worse than open loop) plus a stated rationale. This edge grows with dimension, reaching ~6x fewer evaluations on a 3x3 plant. The behaviour generalizes across four open models, and on a benign plant the LLM offers no advantage, sharpening the boundary. We contribute a reproducible benchmark delimiting when open LLMs help in control tuning: not as optimizers, but as a sample-efficient, interpretable structural prior.

Correctness and readability are key measures of code quality, respectively ensuring functional fidelity and ease of comprehension. While most existing research focuses on improving the correctness of large language models~(LLMs) generated codes, readability remains under-addressed. Enhancing readability through targeted control is challenging due to its subjective nature. In this article, we employ representation engineering~(RepE) as the targeted control method given its characteristics of low data dependency and low computational cost. Prior work on RepE has primarily focused on the targeted control for a single task, but improving the code readability requires the control across multiple tasks. Accordingly we proposes the multitask RepE framework and theoretically discuss the impact of the multitask steering method on the tradeoff between the code readability and correctness. We further provide comprehensive experiments in support. All the relevant implementations are open-source and available upon request.

Skeleton-based action recognition aims to understand human behaviors from body joint sequences and is especially challenging in the one-shot setting, where only a single labeled exemplar is available for each novel action. A key challenge is learning representations that capture the hierarchical and compositional structure of human motion while aligning effectively with high-level action semantics under extreme data scarcity. Existing approaches, largely based on Euclidean embeddings and low-level motion cues, struggle to model the tree-like organization of skeleton data, limiting cross-modal alignment and generalization to unseen action categories. We propose SkelHCC, a unified skeleton hyperbolic CLIP-driven cache adaptation framework for one-shot skeleton-based action recognition. SkelHCC introduces an Explicitly Hierarchical Hyperbolic CLIP (EH-HCLIP) module that embeds skeleton sequences and action language into a shared hyperbolic space. By leveraging the negative curvature and exponential volume growth of hyperbolic geometry, EH-HCLIP naturally encodes the joint-part-body hierarchy of human anatomy and yields structurally consistent cross-modal representations. To support efficient one-shot adaptation, SkelHCC further integrates a training-free LLM-guided Multi-granularity Voting Cache (LMV-Cache) for context-aware inference. Experiments on NTU RGB+D 60, NTU RGB+D 120, and PKU-MMD demonstrate that SkelHCC consistently outperforms state-of-the-art methods.

Deep networks often exhibit a preference for "simple" solutions, and such a simplicity bias is widely believed to play a key role in generalization. Yet a broadly applicable, quantitative measure of simplicity remains elusive. We introduce polynomial representations as a distribution-aware, low-dimensional surrogate for neural functions: we approximate a network's predictive behavior along data-dependent interpolation paths using orthogonal polynomial bases, yielding a compact functional representation. We show that the effective degree of this representation serves as a practical simplicity metric that is predictive of generalization across tasks and architectures, and consistently outperforms existing generalization proxies such as sharpness. Finally, polynomial representations naturally yield a differentiable simplicity regularizer, which consistently improves generalization in image and text classification, fine-tuning contrastive vision-language models, and reinforcement learning.

Pretrained foundation models have become an important basis for end-to-end autonomous driving. In contrast to vision-language models pretrained primarily on static image-text pairs, video generative models capture temporal dynamics and motion priors that are naturally suited for driving. We present DriveWAM, a driving world-action model that adapts a pretrained video diffusion transformer into an autoregressive video-action policy. DriveWAM organizes video and action streams into a unified temporal token sequence and trains them under a joint flow-matching objective, preserving the pretrained video-generation architecture while adapting its large-scale video priors to action generation. To incorporate high-level scene understanding, we introduce scene-evolving driving guidance, where a frozen VLM produces chunk-specific semantic intent to guide video-action generation. To keep long-horizon rollout bounded, we further introduce selective KV memory, which maintains bounded modality-aware video and action memory pools through relevance-redundancy cache selection at inference time. Experiments on NAVSIM and the PhysicalAI-Autonomous-Vehicles benchmark show that DriveWAM achieves strong planning performance, and a data-scaling study from 4k to 100k driving clips further confirms the scaling potential of world-action modeling for end-to-end autonomous driving.

Current motion-controlled image-to-video generation models rigidly follow user-provided trajectories that are often sparse, imprecise, and causally incomplete. Such reliance often yields unnatural or implausible outcomes, especially by missing secondary causal consequences. To address this, we introduce MotiMotion, a novel framework that reformulates motion control as a reasoning-then-generation problem. To encourage causally grounded and commonsense-consistent interactions, we leverage a training-free vision-language reasoner to refine image-space coordinates of primary trajectories and to hallucinate plausible secondary motions. To further improve motion naturalness, we propose a confidence-aware control scheme that modulates guidance strength, enabling the model to closely follow high-confidence plans while correcting artifacts under low-confidence inputs with its internal generative priors. To support systematic evaluation, we curate a new image-to-video benchmark, MotiBench, consisting of interaction-centric scenes where new events are triggered by motion. Both VLM-based evaluation and a human study on MotiBench demonstrate that MotiMotion produces videos with more plausible object behaviors and interaction, and is preferred over existing approaches.

Different age-related regulations have been proposed to protect minors from harmful content and interactions online. Automated age estimation is central to enforcing such regulations, and vision-language models (VLMs) achieve state-of-the-art performance on this task. However, we find that the zero-shot nature of VLM-based age estimation produces an unexpected side effect we call the identity shortcut: Instead of estimating age from visual features, VLMs tend to identify the depicted person and infer their age from memorized knowledge. This phenomenon leads to substantially incorrect predictions when non-celebrities are misidentified as celebrities. It also produces deceptively high robustness to noise and adversarial perturbations on celebrity images, which dominate popular benchmarks. To mitigate this, we propose an activation steering method that suppresses the shortcut by intervening on the hidden states of the VLM. This method improves age estimation accuracy for both memorized and unseen identities, reducing mean absolute error by up to 25% across popular benchmarks.

Large language models (LLMs), trained on vast datasets, encode extensive real-world knowledge within their parameters, yet their black-box nature obscures the mechanisms and extent of this encoding. Surrogate modeling, which uses simplified models to approximate complex systems, can offer a path toward better interpretability of black-box models. We propose a surrogate modeling framework that quantitatively explains LLM-encoded knowledge. For a specific hypothesis derived from domain knowledge, this framework approximates the latent LLM knowledge space using observable elements (input-output pairs) through extensive prompting across a comprehensive range of simulated scenarios. Through proof-of-concept experiments in medical predictions, we demonstrate our framework's effectiveness in revealing the extent to which LLMs "perceive" each input variable in relation to the output. Particularly, given concerns that LLMs may perpetuate inaccuracies and societal biases embedded in their training data, our experiments using this framework quantitatively revealed both associations that contradict established medical knowledge and the persistence of scientifically refuted racial assumptions within LLM-encoded knowledge. By disclosing these issues, our framework can act as a red-flag indicator to support the safe and reliable application of these models.

Despite rapid advances in large language models (LLMs), statistical reasoning remains underrepresented in existing LLM benchmarks, which often do not reflect the layered, proof-driven nature of real statistical practice. To address this gap, we introduce \textbf{StatEval}, the first large-scale benchmark for statistical reasoning across curricular and research-level settings. StatEval includes over 100,000 curated problems, with 20,000+ foundational questions spanning undergraduate and graduate curricula and 80,000+ research-level proof tasks extracted from leading statistical journals. To construct StatEval, we develop \textbf{TRACE} (Topology and Reasoning-Aware Context Extractor), a multi-agent pipeline with human-in-the-loop validation that converts unstructured academic texts into self-contained theorem-level reasoning tasks. We also propose an Adaptive Process-Based Scoring Pipeline for complex statistical proofs, enabling fine-grained evaluation beyond final-answer matching. Experiments show that while LLMs perform reasonably on foundational tasks, they struggle with rigorous research-level reasoning. Beyond evaluation, StatEval serves as a resource for improving reasoning, as retrieval-augmented generation and domain-specific alignment consistently enhance performance. Together, these results establish StatEval as both a benchmark and an infrastructure for advancing statistical reasoning in LLMs.

Watermarking should identify language-model output without degrading quality or limiting verification to the model provider. Multilingual deployment makes this harder because morphology, segmentation, and script change where watermark evidence can enter naturally. We introduce LUNA, a linguistically adaptive watermark that combines model-free detection with single-token non-distortion under the standard random-key model. LUNA estimates normalized next-tag entropy from part-of-speech contexts in an external corpus and uses it to set the depth of a non-distortionary binary tournament sampler; the detector reconstructs the same schedule from text, a tokenizer, a tagger, and a secret key. We evaluate six typologically diverse languages and two domains against eight primary baselines. LUNA attains an AUROC of 0.9959 and the lowest mean absolute median perplexity shift of 0.045 across the twelve settings; its 95% bootstrap interval [0.022, 0.073] lies below all baseline intervals. LUNA also records the lowest mean Self-BLEU, Distinct-1, surprisal, and entropy shifts. It is the only method that simultaneously achieves AUROC > 0.99 and an absolute median perplexity shift below 0.1 in a majority of settings, reaching this regime in 9 of the 12 settings while no baseline reaches it in more than 2. Our code is available at: https://github.com/Shinwoo-Park/luna_watermark

Reward design remains a central challenge in reinforcement learning (RL). Hand-crafted rewards are often difficult to specify and may lead to suboptimal policies, while learned rewards from preferences can suffer from inefficiency and unstable training. Inspired by the dual nature of human learning explored in cognitive science, we decompose rewards into two complementary components: Formal Rewards (FR), explicitly designed based on task knowledge, and Residual Rewards (RR), learned from observations to capture implicit and nuanced preferences. Based on this decomposition, we propose CoRe, a hybrid framework that integrates FR and RR with vision-language models (VLMs) feedback to achieve preference-aligned policies without human involvement. Our contributions are twofold: (1) We propose a Formal Reward Module (FRM) that leverages VLMs to iteratively design and optimize FR based on task knowledge and preference feedback, enabling the continual improvement of policy during training; (2) We introduce a Residual Reward Module (RRM) that learns RR from video-level preference by employing VLMs to generate preference labels and capturing nuanced rewards that complement FR, ensuring alignment with human intent. Through the synergy of FRM and RRM, CoRe enables the automatic construction of reliable rewards that are efficient and preference-aligned. Extensive experiments demonstrate that CoRe outperforms existing approaches in terms of policy learning effectiveness and efficiency on ten robotic manipulation tasks in simulation and five real-worlds. Videos can be found on our project website: https://core-2026.github.io/

In human visual perception, uppercase lettering serves as a natural salience cue that captures attention within lowercase text. In this paper, we present a systematic empirical characterization study revealing that Large Language Models (LLMs) exhibit an analogous property: letter casing modulates internal attention allocation. Through analysis across 13 models, nine LLMs and four Vision-Language Models (VLMs), with diverse tokenization schemes, we show that formatting target information in alternating or uppercase against a lowercase context concentrates attention on those textual spans. In text this effect is universal, holding across every evaluated non-reasoning model. We frame it as a previously under-explored latent property of pretrained transformers rather than a prescriptive method. Our investigation reveals a central attention-performance divergence: while this "casing effect" robustly shifts attention, its impact on downstream accuracy is non-trivial, increased concentration does not inherently improve task accuracy and, in high-entropy contexts like alternating case, can degrade it. We further identify a boundary condition: the deliberative "thinking" phase in reasoning models acts as a semantic buffer that mitigates typographic sensitivity in text. Extending the study to VLMs, we find the effect transfers partially: the same prompt-side casing reorganizes cross-modal attention along two coupled axes, predominantly a macroscopic disengagement from the image toward the text prompt, and secondarily a concentration of the residual visual attention on the target region. By isolating casing as a zero-shot mechanism for attention steering that requires no model access or fine-tuning, we provide a new foundational understanding of how pretraining internalizes typographic emphasis.

Diffusion Language Models (DLMs) offer a compelling alternative to autoregressive (AR) generation by enabling bidirectional context and iterative refinement. However, their reliability under natural input noise and adversarial attacks remains under-explored. To address this, we systematically evaluate DLM robustness and calibration against AR baselines, using two parameter-matched pairs (LLaDA-8B vs. LLaMA-3-8B and Dream-7B vs. Qwen2.5-7B) across 32 natural perturbation conditions, adversarial gradient probes, and mechanistic hidden-state analyses. This paired design effectively isolates architecture-intrinsic properties from weight-dependent behaviors. We find a nuanced robustness profile: while highly stochastic DLM loss landscapes naturally resist gradient-based adversarial suffixes, they provide no guaranteed defense against natural noise, proving that everyday robustness is weight-dependent rather than inherently architectural. Furthermore, DLMs exhibit systematic overconfidence, presenting a practical deployment hazard. Most crucially, mechanistic probing reveals that all models perfectly encode input corruption, isolating behavioral fragility entirely to a decoder routing failure. Consistent with this diagnosis, we show that surface-level prompt patching fails to improve over noisy baselines. Ultimately, DLM robustness cannot be patched on; it must be fundamentally integrated into the iterative decoding loop.

Multimodal Large Language Models (MLLMs) have achieved remarkable success across diverse expert-level tasks, but they still struggle with fundamental abilities that humans naturally develop through continuous observation of the real world, such as spatial perception and dynamic reasoning. Recent studies have recognized this gap and introduced dedicated benchmarks to evaluate the spatial-temporal capabilities of MLLMs. However, existing benchmarks mostly focus on static scenes or require exact quantitative predictions, leaving intuitive reasoning from temporal cues largely underexplored. In this paper, we introduce the Visual Spatial-Temporal Reasoning Benchmark (ViSTR-Bench), a novel evaluation suite designed to systematically assess whether MLLMs can perform qualitative reasoning from continuous visual cues in dynamic scenes. Guided by the principles of temporal emphasis, reasoning orientation, and qualitative evaluation, ViSTR-Bench establishes a comprehensive four-dimensional evaluations covering Motion Perception, Spatial Relations, Outcome Prediction, and Physical Dynamics. The benchmark comprises 15 distinct subtasks and 1,340 high-quality video question-answer pairs spanning diverse tabletop, indoor, and outdoor scenarios. Extensive evaluations of a broad spectrum of state-of-the-art proprietary, open-source, and specialized spatial MLLMs reveal that, despite their strong general video understanding capabilities, current models still face substantial bottlenecks in complex spatial-temporal reasoning and remain far below human performance.

While recent Vision-Language-Action (VLA) models show promise toward generalist manipulation policies, they struggle with long-horizon tasks due to their Markovian nature-relying solely on current observations. Hierarchical dual-system methods address this but suffer from a gap between high-level planning semantics and low-level execution kinematics. We introduce Cortex, a bidirectionally aligned embodied agent framework with a customized planning interface that conveys executable and tractable subtask plans from high-level VLM to low-level VLA. Specifically, we standardize manipulation subtasks into 32 canonical skill primitives and inject tractability principles, such as representative object attributes and improved trajectory reachability, into the data generation pipeline. This enables automatic annotation of over 4k hours of open-source video data and generation of 30 hours of simulation data. We further devise an event-balanced sampling strategy to construct training data for fine-tuning the framework to better handle planning ambiguity during subtask transitions, enhanced by carefully designed harness engineering from task contexts to skill constraints during inference. Both open-loop VLM and closed-loop system evaluations demonstrate Cortex's efficacy, e.g., it outperforms monolithic baselines by 3.1% on Libero-long and 4.1% on RoboTwin. Notably, Cortex's generalist VLM enables zero-shot completion of unseen real-world long-horizon tasks, such as multi-stage chemistry experiments, by simply combining with a fine-tuned VLA-a capability infeasible through VLA fine-tuning alone.

Reinforcement learning (RL) has become a key component in modern large language models, yet the rollout stage remains the key bottleneck in RL training pipelines. Although Multi-Token Prediction (MTP) offers a natural solution to accelerate rollouts through speculative decoding, many studies have observed that MTP acceptance rates degrade significantly during RL training, leading to limited speedup performance. To address this bottleneck, we present Bebop, a systematic study of MTP in LLM post-training, and offer practical recipes to integrate MTP into large-scale RL pipelines. First, we reveal that the MTP acceptance rate is fundamentally bounded by the fluctuation of model entropy, which demonstrates a clear negative linear relationship with the rise of entropy in the RL stage. Second, we show that probabilistic rejection sampling largely alleviates the disturbance introduced by entropy in RL compared to greedy draft sampling. We further identify that the conventional MTP training objectives (cross-entropy or KL) are suboptimal in such settings, and therefore we propose a novel end-to-end TV loss that directly optimizes multi-step rejection sampling acceptance rate, yielding ~10% acceptance rate improvements, achieving up to 95% acceptance rates and up to 25% extra inference throughput gains across mathematical reasoning, code generation, and agentic tasks. Third, we test various online MTP training strategies during RL and show that pre-RL MTP training with e2e TV loss and rejection sampling achieves a consistent acceptance rate and speedup throughout the entire RL, eliminating the need for costly online MTP updating. We provide extensive experiments and analysis that validate our findings. Experimental results show our method achieves up to 1.8x end-to-end acceleration in async RL training of Qwen3.5, Qwen3.6, and Qwen3.7 models.

Small open-source code models that power IDE autocomplete still emit hallucinated Fill-in-the-Middle (FIM) completions: syntactically natural calls to methods, parameters, variables, and imports that do not exist in the surrounding project. Existing mitigations either require per-language execution sandboxes that do not apply at mid-keystroke or preference-optimisation pipelines that need large human-labelled corpora. We propose an execution-free alternative: use frontier code models to synthesise plausible-but-wrong completions as hard negatives, then leverage the contrast between these synthetic hallucinations and the ground-truth developer edit as a supervised fine-tuning signal. Our pipeline scrapes multilingual FIM contexts from public GitHub across eight languages and asks a panel of three frontier generators to produce one hard negative per context for each of four hallucination types drawn from the Delulu taxonomy, a Docker-verified multilingual FIM hallucination benchmark, yielding a paired chosen/rejected dataset. Fine-tuning Qwen2.5-Coder-7B-Instruct on a 100K-row curated subset lifts Delulu exact match by +18.8 points and edit similarity by +0.22 on every language and every type, while also improving every HumanEval-Infilling split and every SAFIM subset. The same recipe at 3B lifts Delulu by +12.8 EM with a small, characterised general-FIM trade-off. Five-axis ablations (size, type mix, language coverage, base-model family, and a difficulty-aware fool rate) plus a head-to-head SFT vs. DPO/ORPO comparison map which design choices drive the gain. We release the full pipeline source code -- generation, fool-rate LLM judging, curation, and the FIM fine-tuning recipe -- so that the experiments in this paper can be reproduced end-to end on any permissively licensed corpus.

Traffic prediction plays a central role in intelligent transportation systems (ITS) by supporting real-time decision-making, congestion management, and long-term planning. However, many existing approaches face practical limitations. Most spatio-temporal models are trained on centralized data, rely on numerical representations, and offer limited explainability. Recent Large Language Model (LLM) methods improve reasoning capabilities but typically assume centralized data availability and do not fully capture the distributed and heterogeneous nature of real-world traffic systems. To address these challenges, this study proposes FedLLM (Federated LLM), a privacy-preserving and distributed framework for explainable multi-horizon short-term traffic flow prediction (15-60 minutes). The framework introduces four key contributions: 1) a Composite Selection Score (CSS) for data-driven freeway selection that captures structural diversity across traffic regions 2) a domain-adapted LLM fine-tuned on structured traffic prompts encoding spatial, temporal, and statistical context 3) FedLLM framework, that enables collaborative training across heterogeneous clients while exchanging only lightweight LoRA adapter parameters, 4) a structured prompt representation that supports contextual reasoning and cross-region generalization. The FedLLM design allows each client to learn from local traffic patterns while contributing to a shared global model through efficient parameter exchange, reducing communication overhead and keeping data private. This setup supports learning under non-IID traffic distributions. Experimental results show that FedLLM achieves improved predictive performance over centralized baselines, while producing structured and explainable outputs. These findings highlight the potential of combining FL with domain-adapted LLMs for scalable, privacy-aware, and explainable traffic prediction.

Humans understand and interact with the real world by relying on diverse physical feedback beyond visual perception. Motivated by this, recent approaches attempt to incorporate physical sensory signals into Vision-Language-Action models (VLAs). However, they typically focus on a single type of physical signal, failing to capture the heterogeneous and complementary nature of real-world interactions. In this paper, we propose MoSS, a modular sensory stream framework that adapts VLAs to leverage multiple sensory signals for action prediction. Specifically, we introduce decoupled modality streams that integrate heterogeneous physical signals into the action stream via joint cross-modal self-attention. To enable stable incorporation of new modalities, we adopt a two-stage training scheme that freezes pretrained VLA parameters in the early stage. Furthermore, to better capture contact interaction dynamics, we incorporate an auxiliary task that predicts future physical signals. Through extensive real-world experiments, we demonstrate that MoSS successfully augments VLAs to leverage diverse physical signals (i.e., tactile, force, and torque), integrating multiple signals to achieve synergistic performance gains.

Growing attention to intelligent agents has put a spotlight on one of their central capabilities: planning. Early attempts to leverage large language models (LLMs) for planning relied on single-shot plan generation, followed by hybrid approaches that coupled LLMs with limited external search. These methods, unsound and incomplete by their very nature, often require substantial resources without yielding better solutions on unseen problems. As the limitations of LLMs become clearer, recent work has shifted toward using them at solution construction time -- generating symbolic solvers for a family of problems that can be verified and then used efficiently at inference time. This trend reflects the growing need for agents that are both reliable and resource-efficient. It also offers a path towards generating maintainable planners with minimal dependence on language models at inference time. In this paper, we argue that this shift reflects a broader realignment of the planning field in the LLM era. We examine three major categories of planner-generation methods, discuss their current limitations, and outline research steps towards a more reliable and efficient LLM-based generation of planners.

Current speech synthesis struggles with code-switching, which mixes a foreign language phrase into a primary language utterance, causing the phrase to be spoken with the primary language's accent rather than its native one. We propose Phrase-Localized Language-Contrastive Guidance (LCG), a training-free inference framework that restores a native accent to code-switched phrases in cross-lingual text-to-speech. LCG replaces the single language guidance applied across the whole utterance with a separate guidance for each region, so each part is guided by its own language. To choose where to apply this localized guidance, we propose a self-attention probing technique that finds the phrase boundaries without external alignments. Together, these components generate speech in which each region carries the accent of its own language, requiring no fine-tuning or auxiliary models. Across diverse language pairs, LCG robustly increases the nativeness of the code-switched phrase while suppressing accent leakage, and preserving overall speaker identity and naturalness.

Large Language Models (LLMs) have demonstrated remarkable performance across diverse tasks, from text summarization to question answering. Despite these capabilities, their black-box nature obscures internal decision-making processes. Mechanistic interpretability (MI) aims to address this by reverse-engineering neural networks into human-understandable algorithms. Current MI approaches for LLMs typically follow a two-stage paradigm: first identifying important components (circuit discovery), where components are typically individual nodes such as an attention head or feedforward neuron, and second determining the role they play in a certain task (functional interpretation). However, this sequential approach overlooks a fundamental insight: a component's importance and its functional role are inherently codependent. Unifying these stages presents two key challenges: (1) functional roles are often tied to specific nodes or components, limiting generalization, and (2) their identification relies on subjective interpretation rather than quantifiable metrics. To address these challenges, we propose S^3martCirc (Self-supervised Smart Circuit Discovery), a unified framework that simultaneously discovers circuits and interprets functionality. S^3martCirc abstracts node behavior into two general computational roles that generalize across tasks and defines a quantitative metric for assigning them, enabling importance and functional role to be discovered jointly rather than in sequence. Extensive experiments show that our framework outperforms existing methods in circuit discovery.

Test-time reinforcement learning (TTRL) enables language models to self-evolve at inference time without labeled feedback. Existing methods rely on answer voting and therefore do not extend naturally to open-ended generation, where valid responses cannot be mapped to a shared canonical answer. Without external reward models or stronger judges, adaptation must instead construct reliable rewards from the model's own outputs. We introduce SERPO (Self-Evolving Rubric Policy Optimization), which replaces answer voting with a closed loop that co-evolves response evidence, query-specific rubrics, and policy parameters. Good-Normal-Bad (G-N-B) response evolution organizes maximally separated rollouts into ordered archives; rubric evolution retains criteria that discriminate these archives; probabilistic criterion scoring converts verdict-token likelihoods into reward signals; and policy evolution optimizes the actor with the resulting signals. New actor rollouts then refresh both the archives and rubrics, closing the three-way evolution loop. Across two model configurations, two in-domain benchmarks, and four OOD benchmarks, SERPO improves HealthBench and ResearchQA by up to 20.63 and 20.31 points over the corresponding base models, raises the six-benchmark macro-average by up to 8.06 points, and supports OOD transfer and continued cross-benchmark evolution.