每天知道
你的领域
出了什么

Uncertainty quantification (UQ) for large language models (LLMs) aims to provide reliable measures of predictive confidence, yet current methods are often unstable under meaning-preserving perturbations. Semantically equivalent paraphrases can induce substantial variability in predictive confidence, even for methods with formal guarantees, such as conformal prediction. To address this issue, we propose a paraphrase-aware UQ framework robust to semantic rewordings. Our approach trains a lightweight proxy model on LLM hidden states and aggregates its predictions across paraphrases to construct label-wise nonconformity scores. Under score exchangeability, conformal calibration retains marginal coverage. This guarantee can also hold under test-only rewording, provided that the paraphrase pipeline satisfies an additional distributional alignment condition. We evaluate three settings (normal, fully reworded, and semi-reworded) which apply rewording to neither dataset, both calibration and test datasets, or only the test dataset, respectively. Across seven multiple-choice QA benchmarks and multiple model families, our method produces compact prediction sets with empirical coverage generally near the nominal target, even in the semi-reworded setting. Ablation studies show that the learned proxy accounts for most of the reduction in set size, while paraphrase-augmented training and inference-time aggregation improve stability under rewording. Code is available at https://github.com/Raina-Xin/PA_Score.

Reinforcement learning with verifiable rewards (RLVR) has been shown to improve the reasoning capability of large language models (LLMs) across diverse reasoning tasks. However, group-based RLVR methods, such as GRPO, assign a uniform advantage to all tokens within rollouts of the same outcome. While existing works refine credit assignment of GRPO based on local signals such as token locations or entropy, they often fail to capture the global semantic novelty of a reasoning behavior relative to the current policy. In this work, we propose a hierarchical credit assignment approach for group-based RLVR methods, called HarA, which identifies and encourages semantically novel reasoning behaviors during RLVR. HarA represents each sampled rollout as a distribution over the hidden states and locations of tokens, and computes the Fused Gromov-Wasserstein (FGW) barycenters of all rollouts with the same outcome, capturing the internal reasoning patterns in the latent space under the current policy. The semantic novelty of a reasoning element can then be measured by its contribution to the FGW distance between the current rollout and the barycenter. While solving the FGW formulation is expensive, we introduce an anchor-guided linearization that turns it into a Wasserstein formulation solvable via the Sinkhorn algorithm efficiently. By reweighing token-level advantage of group-based RLVR methods based on the novelty signals, HarA highlights novel reasoning behaviors at flexible granularities to encourage fine-grained LLM exploration. Extensive experiments across three group-based RLVR methods show that our plug-and-play method effectively enhances the exploration of LLMs, outperforming existing methods across diverse reasoning benchmarks.

Large language models (LLMs) are increasingly deployed in safety-critical settings, yet their black-box nature makes it difficult to provide formal guaranties about their behavior. Existing verification approaches rely primarily on empirical probing and testing, leaving open the question of how to reason rigorously about general-purpose trans- former architectures. In this work, we establish a principled bridge between transformers and weighted automata, a classical model from formal language theory. This connection enables us to transfer verification tools from automata the- ory to the analysis of LLMs. Our contributions are twofold: First, we develop a formal correspondence between transformer architectures and weighted automata over reals, showing how distributional properties of LLMs can be captured within this framework. Second, we introduce an identity testing algorithm for weighted automata that provides a statis- tical method for distinguishing whether two stochastic models define the same distribution up to a tolerance threshold. This work provides the first formal bridge between modern neural se- quence models and classical automata theory, clarifying both the poten- tial and the computational challenges for rigorous LLM verification.

Answer correctness is encoded as a recoverable geometric direction in the hidden states of language models. We show that the mean displacement from incorrect to correct answer representations, computed at approximately 70% of model depth from fifty labeled examples with no parameter updates, yields a scoring direction that outperforms zero-shot log-probability scoring by up to +32.0 percentage points on factual benchmarks (ARC-Challenge and MMLU) and by +38.1 to +51.8 percentage points on TruthfulQA, across five models spanning 1B to 8B parameters in three architecture families (Llama, Qwen, Gemma). The method requires one forward pass and one dot product per candidate; no generation is performed at inference. Applied as a hallucination detector on individual (question, answer) pairs, the recovered direction achieves 0.693~AUROC versus 0.578 for log-probability scoring. We additionally find that correctness directions for factual reasoning, domain knowledge, and calibrated truthfulness are near-orthogonal in representation space, revealing that language models allocate geometrically independent subspaces to qualitatively distinct notions of correct answer, with architecture-dependent variation in the degree of separation. This structure explains the observed transfer pattern---the direction calibrated on factual questions transfers within task type but not across it---and suggests that LLM calibration failures may reflect a routing problem: the model's internal representation contains more correctness signal than its output behaviour exploits.

Reinforcement fine-tuning (RFT) is increasingly used in applications where large language models (LLMs) interact with humans and other agents. Here we use social deduction games to study how RFT changes LLMs' social behaviour. We let fine-tuned and base LLM agents play hidden-role games that require hidden-state inference, social reading and vote steering. Our results show that LLM agents do not reliably acquire social-deduction ability by directly optimizing terminal win--loss outcomes, suggesting that final game results provide a sparse and noisy signal for socially interactive learning. However, RFT is particularly effective at improving social reading, including tasks that require agents to infer hidden roles from public discussion, update beliefs over time and predict other agents' future decisions. We further show that RFT can also improve social influence, including tasks that require agents to steer votes, team approvals and collective decisions, although these gains depend more strongly on behaviourally specific rewards and structured interaction settings. Finally, we show that LLMs' ability to play social deduction games can be further improved through multi-agent social-cognitive reinforcement fine-tuning, which combines social-reading and social-influence signals during same-side multi-agent training. These learned behaviours also receive more favourable human evaluations of strategic competence, persuasiveness and social usefulness. Together, these results enrich our understanding of how RFT changes LLMs' social behaviour and provide a step toward a behavioural learning theory for machine social intelligence.

Recent studies on reinforcement learning (RL) report seemingly conflicting evidence about large language model (LLM) reasoning. Training on mathematics can improve performance in other domains, yet gains in Pass@1 can coincide with lower Pass@$N$ than the base model. This raises a fundamental question: does RL expand an LLM's reasoning boundary, or merely reweight its existing reasoning space? We revisit these phenomena across Qwen and Gemma model families, showing both cross-domain gains and forgetting, while coverage at large sampling budgets increases on some tasks and decreases on others. Detailed analysis of solution traces before and after RL indicates a shift in the reasoning strategies the model employs, motivating a two-stage autoregressive policy model that separates strategy selection from problem-specific execution. Within this framework, we prove how RL's implicit bias reshapes strategy preferences, allowing gains on some tasks while suppressing strategies required by others. This mechanism can also broaden or narrow coverage at a given sampling budget even without expanding strategy support. We further provide theoretical justifications for log-sigmoid and log-linear scaling laws in RL compute, and evaluate their predictive power. Together, these results connect changes in strategy selection to cross-domain transfer, reasoning coverage, and compute scaling.

Efficient serving of Mixture-of-Experts (MoE) models is challenging due to large expert parameters, input-dependent expert activation, and dynamic workloads. Expert parallelism distributes expert computation across GPUs, while attention-FFN disaggregation (AFD) separates attention and feed-forward computation into independent worker pools. However, we observe that a naive AFD implementation could make expert load imbalance more harmful: once FFN computation becomes an independent pipeline stage, overloaded experts directly slow the FFN stage and degrade end-to-end serving performance. To solve this problem, we present AFORE, a timely expert reconfiguration system for AFD-based MoE serving. AFORE exploits two architectural properties of AFD. First, AFD exposes the expert-token distribution of upcoming microbatches before they reach FFN execution, enabling placement decisions based on near-future demand instead of stale historical profiles. Second, AFD creates a pipeline window in which expert migration for a target microbatch can be overlapped with the computation of preceding in-flight microbatches. AFORE formulates expert reconfiguration as a microbatch-aware scheduling problem and uses a migration-aware scheduler to decide when and which experts to migrate. AFORE further implements lightweight demand prefetching and NVLink-based GPU-GPU expert migration to reduce reconfiguration overhead. Evaluation on a 110B-parameter MoE model across four dynamic workloads shows that AFORE improves output throughput by 10.1-17.6% and reduces P95 inter-token latency by 7.1-9.5% compared with the strongest competing baseline. Compared with static placement, AFORE improves throughput by 29.8% on average and reduces P95 inter-token latency by 18.2% on average. Migration profiling further shows that AFD pipeline overlap can fully hide expert-migration latency.

Recent advances in Large Language Models (LLMs) have demonstrated remarkable capabilities in web front-end execution, with browser-based game generation emerging as a particularly prominent frontier. While previous efforts frequently rely on complex multi-turn workflows or focus on static game evaluation benchmarks, this work targets direct end-to-end real-world game synthesis driven by coding agents. However, generating complex games directly from sparse user queries often forces coding agents to make underspecified assumptions, yielding incomplete mechanics, disconnected gameplay flows, and limited visual aesthetics. To resolve this issue, this paper presents GameGo, a scalable framework that systematically transforms brief game seeds into comprehensive Product Requirements Documents grounded in industry game-development practices. To retain core gameplay constraints without restricting design exploration, GameGo uses task-specific dynamic compression to maximize information density while preserving instruction following. Based on this pipeline, GameGoData is constructed with 55,060 development trajectories across 2D, 2.5D, and 3D games, alongside GameGoBench, a benchmark comprising 124 diverse game queries. Training GameGoCoder on GameGoData yields a model that outperforms matched baselines and is comparable to frontier models across gamedev benchmarks. All code, datasets, and models will be made publicly available.

Artificial intelligence has advanced individual radiotherapy tasks, yet these capabilities remain separated across clinical stages, software environments and data modalities. This fragmentation contrasts with the longitudinal radiotherapy workflow from treatment decision-making through follow-up. Here we present RadOnc-Agent, an agentic artificial-intelligence framework that formalizes radiotherapy into four clinical phases and provides 26 callable functions through a conversational interface. A large-language-model controller maps clinical intent to schema-constrained calls, preserves patient and workflow context, and routes requests to specialist services. We evaluated system execution using 2,600 single-function requests (7,800 repeat executions), 200 prespecified synthetic cross-stage scenarios spanning four phases (600 executions), and 120 workflow instances from 60 de-identified patient records (360 clean executions) representing decision-to-planning and planning-to-adaptation. RadOnc-Agent selected the intended function in 98.79% of single-function executions, completed 96.50% of scripted cross-stage workflows, and completed 96.67% of real-patient workflow executions. In comparative ablations, removing longitudinal state reduced cross-stage completion from 96.50% to 84.00%, while disabling schema and identity validation increased mismatched backend dispatch from 0% to 95.28% in a replay/test evaluation. These findings establish the technical feasibility of an LLM-orchestrated architecture for coordinating heterogeneous radiotherapy capabilities and information across longitudinal workflows; they do not establish clinical correctness, clinical utility or prospective benefit.

Prefill-decode disaggregation is becoming a common architecture for LLM serving because it separates two phases with distinct execution patterns and SLO objectives. Existing systems typically combine a fixed prefill/decode worker ratio with request routing across workers. However, real-world workloads exhibit both short bursts and sustained shifts in the prefill-to-decode demand ratio. As a result, a configuration that is well provisioned at one time may quickly become mismatched, causing latency SLO violations even when idle capacity exists elsewhere. Existing autoscaling mechanisms can add capacity, but they react slowly, require spare GPUs, and do not directly address short-timescale phase imbalance. We present FluidPD, a P/D-disaggregated serving system that provides SLO-aware in-place elasticity. FluidPD introduces two complementary mechanisms. FluidToken handles transient imbalance by offloading a bounded portion of prefill computation to decode workers when decode-side slack is available. FluidRole handles sustained imbalance by reassigning running workers between prefill and decode roles in place, avoiding model reload and engine restart. Both mechanisms are guided by lightweight pressure indices that expose prefill and decode-side resource pressure before they appear as SLO violations. Across production Azure trace workloads, FluidPD improves overall SLO attainment over static SGLang by up to 94.6 percentage points, demonstrating that SLO-aware in-place P/D elasticity improves service quality without provisioning additional workers.

Reinforcement learning with verifiable rewards improves reasoning, while the allocation of learning signal shapes which solutions remain accessible under repeated sampling. Group-relative objectives assign equal advantages to equally rewarded responses, making aggregate credit proportional to sampled mode frequency. We introduce Exploration-Preserving Policy Optimization (ExPPO), a lightweight advantage-shaping rule that redistributes credit using prompt-relative, length-normalized response surprisal and prompt pass rate. ExPPO combines bounded shaping with shared normalization to preserve verifier polarity and approximately maintain each prompt group's total absolute sequence-advantage mass. Our analysis characterizes response-level credit allocation alongside sampled mode updates, deriving local conditions for gains in entropy and correct-mode discovery. Experiments show improved in-domain and out-of-domain reasoning coverage, higher aggregate response accuracy, and strong coverage at large sampling budgets. A controlled multi-answer evaluation further demonstrates increased correct-mode yield and gains in diversity among verified-correct responses. Code is available at https://github.com/jinhangzhan/ExPPO

Developing socially intelligent AI remains heavily dependent on human-annotated data, limiting the scale and breadth of social understanding models can acquire. Methods that derive training signals from unlabeled data offer a path beyond this dependence, but social predictions lack the verification oracles available in mathematics and coding. Moreover, core social targets such as affect, intent, preference, and pragmatic meaning are often ambiguous. The same behavior can support multiple plausible interpretations, making it difficult to verify which is best supported. To address this challenge, we introduce Reinforcement Learning with Comparative Evidence (RLCE), a reinforcement learning method that learns social understanding from unlabeled training data without constructing rewards from ground-truth annotations. Given distinct answers in a rollout group, RLCE constructs evidence tests that identify observable evidence favoring an answer over another, validates these tests against the input sample, and aggregates test outcomes to determine the best-supported interpretation. Tests are regenerated as the policy produces new answers, enabling them to evolve with the policy. Across four benchmarks spanning affect, pragmatics, communicative intent, and preference, RLCE attains the strongest performance among seven methods that use no ground-truth training labels for rewards, including consensus, policy LLM-judge verification, multimodal co-evolution, and rubric-based rewards. Gains over the strongest baseline reach up to +18.93 points. Analyses further show that RLCE exhibits a larger share of reward variation between correct and incorrect predictions than compared rubric methods, can overturn erroneous policy-derived preferences, and benefits from pairwise test construction, compositional test aggregation, and on-policy test evolution.

Large Audio Language Models (LALMs) are prone to hallucinating and over-relying on text priors when simultaneously presented with audio and text inputs. To mitigate these hallucinations, we propose utilizing the multimodal Direct Preference Optimization (mDPO) objective, which forces the model to ground its generation in the acoustic input by contrasting intact and distorted audio counterparts. We extend this preference learning framework to the audio domain by applying a variety of acoustic perturbations. Evaluating the Qwen2-Audio backbone across the DCASE 2025 Challenge and AH Existence datasets, we demonstrate that extending mDPO to LALMs significantly enhances temporal reasoning in the complex DCASE dataset, and improves performance on basic existence verification in the AH benchmark. We identify temporal reversal, frequency masking, and random noise as the most effective perturbations. Ultimately, our approach achieves an absolute accuracy improvement of 14.0% on the DCASE 2025 dataset and 27.4% on AH Existence.

Inverse materials design is a long-standing goal of computational materials discovery. Generative models for crystalline materials are typically trained to match the distribution of a structure database, while nothing in their training objective points them at specific design goals such as targeted properties. We use group-relative policy optimization (GRPO) to align a generative model based on stochastic interpolants and discrete flow matching with general black-box reward functions through reinforcement learning. Atom types are generated by a discrete flow and the policy gradient of our generalization of GRPO directly acts on the likelihoods of the atom-type transitions, which differentiates our work from previous reinforcement-learning approaches for diffusion and flow-based generative models of crystalline materials. We introduce a reward function that raises the yield of metastable, unique and novel structures (mSUN) from 13.4% for the pretrained model to 45.5% for the reinforced model, as evaluated by a community benchmark. Our reward also improves the performance of a reinforcement learning framework for crystalline materials based on latent denoising diffusion models. At the same time, we find that directly reinforcing atom-type transition likelihoods enables reward exploitation that has to be prevented with explicit guards. The same analysis also exposes a gap in the community metric. Single-element structures in distinct packings are counted as metastable, unique and novel materials and inflate mSUN without yielding any new compounds. A stability claim is only as good as its reference hull. We report every result split by the number of reference phases behind it and argue that benchmarks should do the same.

Knowledge distillation is crucial for deploying vision-language models on resource-constrained devices. However, existing methods typically impose uniform supervision across visual tokens or rely on static token selection, which confuses task-relevant cues with background noise and degrades cross-modal reasoning. To address this limitation, we propose Key Visual Evidence-guided Knowledge Distillation (KVE-KD), a framework that dynamically focuses feature distillation on task-relevant visual tokens identified by the teacher model. Specifically, KVE-KD appoints the final pre-generation textual token as a unified semantic anchor and identifies the target cross-modal fusion layer by analyzing changes in the anchor representation through iterative visual-token contribution removal. Within this layer, KVE-KD ranks visual tokens via the anchor-conditioned attention distribution and selects the most informative visual tokens as key visual evidence with normalized entropy. The key visual evidence subsequently guides focused visual feature distillation, making the student align closely with the teacher's task-relevant visual representations while suppressing irrelevant background information. Extensive experiments on six benchmarks demonstrate that KVE-KD outperforms state-of-the-art cross-modal distillation methods, with particularly pronounced gains on tasks requiring complex reasoning and fine-grained visual understanding. Importantly, these improvements are achieved without introducing any inference-time overhead. The source code is available at https://github.com/zhangjianbin07/KVE-KD.

Text-driven human motion generation has advanced substantially, yet most methods assume instructions are available before synthesis. Interactive applications require responding to new instructions while continuing ongoing actions, such as answering a phone while walking. Existing approaches address streaming generation or simultaneous composition without explicitly combining streaming instruction arrival with independently timed, overlapping actions. We introduce streaming multi-track timeline control and propose TimelineControl to incorporate new instructions alongside ongoing actions. Interval-aware conditioning preserves instruction timing, while causal part-structured representations and part-aware denoising coordinate concurrent actions across body regions. We also construct TimelineMotion, a dataset with overlapping instruction intervals and body-part annotations. Experiments on TimelineMotion and MTT demonstrate improved semantic alignment and temporal adherence over evaluated streaming baselines, including models retrained on the same data. Ablations and human evaluations validate our design, complemented by spatial conditioning and humanoid execution demonstrations. Our code, data and models will become publicly available.

Limited training data diversity constrains generative modeling of 3D human bodies: conservative models remain close to observed examples, whereas exploratory models often violate basic body proportions. We introduce a verifier-guided augmentation framework that uses global and mode-local PCA to generate inexpensive candidates, screens them using correspondence-derived skeletal proportions and body-part geometry, and retrains a diffusion model on accepted candidates. Elastic registration provides both the modal structure used by distributed PCA and the dense anatomical correspondence needed for scalable screening without per-candidate body-model fitting. A blinded human study supports the verifier as a conservative gatekeeper, favoring verifier-accepted over rejected outputs. We evaluate full-pool verifier acceptance separately from the coverage and departure of accepted samples and combine them through EAUC. On 4,498 registered DFAUST surfaces, distributed-PCA augmentation achieves 86.32% acceptance, the highest CP-AUC (0.871), and the highest EAUC (0.752), improving EAUC by 32% over real-only and self-augmented diffusion. These results show that mode-local, verifier-guided proposals broaden diffusion generation while maintaining high agreement with calibrated body measurements.

Generating a boundary representation (B-rep) conditioned on a single image requires faithful reconstruction of geometry, valid topology, and support for complex shapes. We present UniBRep, a geometry-first framework that adapts a pretrained image-to-3D model to generate a feature mesh as a unified intermediate representation. Its surface provides a geometric scaffold, while spatially aligned learned features encode face-separation cues for topology recovery. Dual decoder branches generate the geometry and face-separation features; a geometry- and feature-guided construction pipeline then fits parametric surfaces, recovers boundary curves and connectivity, and assembles an explicit B-rep using a CAD kernel. Recovering topology from mesh regions avoids predefined architectural face-count limits, allowing face count to scale with shape complexity. On the standard DeepCAD benchmark, UniBRep produces valid B-reps for 80.49% of inputs and reduces face Chamfer distance from 0.1096 to 0.0345 relative to CADDreamer. In a matched comparison, UniBRep also outperforms the HoLa public demo across all reported metrics. Further evaluations demonstrate scalability to high-complexity shapes beyond the standard 30-face range, generalization to objects outside the CAD training distribution, and qualitative transfer to real photographs.

Vision-language-action models (VLAs) inherit strong semantic grounding from pretrained vision-language backbones but are typically optimized for predicting actions rather than future observations. They can see and act, but they do not imagine the future before acting. World action models (WAMs) built on video diffusion backbones can imagine but treat language as frozen conditioning on a continuous latent space. Unified models bring these modalities into one architecture, but they either decode autoregressively, one token at a time, or keep video continuous with an auxiliary action head. In this work, we present SUAVE, a Single vocabulary Unified Action-Video modEl in which a masked diffusion transformer generates video and actions conditioned on language, with all three modalities represented as discrete tokens in a shared sequence. Choosing which tokens to mask at inference turns the same network into a world model, a robot policy, or a video-action model. For action-free co-training, the action positions of unlabeled video are filled with mask tokens and excluded from the loss. Simulation and real-world experiments demonstrate two findings. First, a single SUAVE model predicts long-horizon video and acts as a policy, competitive with dedicated world models and specialized action policies on static and dynamic manipulation tasks. On a real robot, our model generates subgoal images and an action chunk spanning one second of motion in 1,030 ms on an RTX 5090 GPU, sustaining closed-loop control at 2.5 actions per second. Second, pretraining on robot video and co-training on human video substantially improves policy performance and zero-shot robustness to distribution shift. Together, these results show that masked diffusion is a practical and versatile foundation for unified video-action modeling.

Video world models aim to preserve scene structure and predict how dynamic objects evolve beyond visual observations. We present Kepler4D, a framework for future video generation through explicit 4D scene state evolution. Given a monocular video, Kepler4D constructs a shared 3D representation of background geometry, object motion histories, coarse spatial supports, and semantic context. Chain-of-Motion summarizes observed motion and uses a vision-language model to select structured speed and heading decisions and decide whether to bound object-center height from below. A deterministic rollout converts these decisions into future object trajectories for inspection and editing before synthesis. We render the evolving proxies into geometric controls for a pretrained video generator, separating coarse object motion from the synthesis of appearance and articulation. Experiments on real-world videos demonstrate that Kepler4D enables controllable object motion and plausible future rollout while preserving scene consistency.

Autoregressive transformers remain comparatively weak for protein sequence and structure generation. We study the role of target representation: amino acid tokens encode residue identities without explicit contextual semantics, while backbone coordinates require a discrete representation in our framework. We introduce two learned latent protein languages. Protein Latent Language (PLL) maps sequences to a 4,096-state contextual alphabet built on a frozen ESM-2 encoder, with one token per residue. Structure Latent Language (SLL) adapts GCP-VQVAE Lite with auxiliary sequence and confidence supervision while retaining decoding to backbone coordinates. We separately pretrain autoregressive transformer models on PLL and SLL tokens using next-token prediction, yielding PLLM and SLLM. Under matched downstream sequence training, PLLM has a fitted compute-scaling exponent of 0.038 versus 0.020 for the amino acid autoregressive model. In unconditional sequence generation, PLLM reduces the fraction of samples below a heuristic 1.5-bit residue-composition entropy threshold by 54% relative to the amino acid model across sampling temperatures. For sequence-to-structure prediction, replacing the original GCP-VQVAE Lite tokenizer with SLL reduces best validation perplexity by 34% under matched training. For long proteins, latent-token sampling is approximately 1,000 times faster than MSA-based AlphaFold2 in our measurements. In backbone generation, SLLM compares favorably with other generative models on diversity and novelty. We also observe early signs that using SLLM's internal token confidence for inference-time sampling can improve sequence-to-structure prediction quality beyond a single decoded sample. These results position learned latent protein languages as a promising substrate for autoregressive transformer scaling and inference-time sampling in protein generation.

A novel discovery is one which is both useful and surprising: a generative model's output is a useful discovery if it has a low probability of being generated (it's surprising) and a high reward (it's useful). Global optimization can directly increase the probability of sampling high rewards but typically requires updating model weights. Such gradient based optimization is expensive and bars using capable closed-source models. Instead, modern search methods for discovery sacrifice the global target, and use evolutionary algorithms with local reward maximizing objectives, permitting the search to focus only on high probability samples. In this paper, we interpret various evolutionary algorithms as approximate Markov Chain Monte Carlo, an optimization-free method to sample from complex distributions. This interpretation allows developing Distribution Matching Evolutionary Algorithms (DME), a class of search methods which sample from a global target distribution without updating weights. Empirically, DME has a higher sample efficiency than existing methods on problems requiring many samples to find a solution.

Text-to-video (T2V) diffusion models can reproduce copyrighted, violent, or explicit content, which motivates concept erasure: removing designated concepts from a pretrained model while preserving its behavior on everything else. Existing T2V erasure methods leave two problems open. Their frame-agnostic suppression can leave isolated frames in which an erased concept resurfaces, a frame-reactivation gap that clip-level averages obscure; and they are usually evaluated with one target concept or category at a time. We propose Frame-Aware Diffusion Erasure (FADE), a multi-concept video unlearning framework. FADE first applies a joint closed-form key/value edit that suppresses all target concepts, then trains per-concept frame-aware low-rank adapters whose strength is gated by the frame index and the denoising timestep to remove residual per-frame leakage. Each adapter is trained with the other targets' prompts as hard negatives, which keeps the concept-specific components of different adapters well separated, and a similarity-based soft router combines the adapters according to the prompt. With 16 concepts (objects, artistic styles, and nudity) erased from a single Wan2.1-T2V-1.3B backbone, FADE reduces the residual accuracy on the object benchmark to 4.9%, against 15.5% for the strongest of eight baselines, while keeping the VBench average within 0.9% of the unedited model. The ranking is unchanged under a VLM judge and a blinded human study, and the advantage over the strongest baseline carries over to prompts that combine several erased concepts, to 30 simultaneously erased celebrity identities, and to Wan2.1-T2V-14B, CogVideoX-2B, and HunyuanVideo-1.5.

Synthetic data are increasingly used as an alternative to sharing sensitive records. However, synthetic data generation does not guarantee privacy, as diffusion models trained or adapted on sensitive data remain susceptible to reconstruction attacks. Moreover, while approaches that use differential privacy (DP), such as DP-SGD, achieve provably private diffusion model training, the repeated gradient clipping and noise injection they require result in significant utility loss. An important limitation of DP-based privacy is that, although it has a provable relationship to reconstruction privacy (RP), that relationship is indirect. RP is defined in terms of limiting how much an adversary's posterior distribution over sensitive data differs from the prior, whereas DP provides guarantees by bounding the sensitivity of outputs to changes in individual records. This indirection is an important source of the utility loss. To address this, we propose a PAC-private diffusion model adaptation to achieve reconstruction privacy. Since PAC-privacy is defined directly with respect to posterior advantage over the prior, it directly implicates RP. To obtain scalable PAC privatization in high dimensions, we first learn a compact data-dependent diffusion model component using LoRA or Textual Inversion, and then calibrate anisotropic Gaussian noise from the covariance of repeated mechanism outputs. Unlike DP-SGD, our method perturbs the learned component only once after optimization, thereby avoiding privacy composition across gradient updates. We evaluate the framework on few-shot concept personalization and full-dataset image synthesis, and show that the proposed approach better preserves subject identity, generation quality, and downstream classification accuracy than DP while achieving the same reconstruction privacy.

Synthesizing realistic graphs at scale is vital when the graphs of interest are large and real-world samples are limited or access-sensitive. Diffusion-based generators have recently driven much of the progress, offering high modeling capacity, but most such methods have quadratic computational complexity and are hence restricted to small-scale networks, currently up to 3k nodes. Existing non-quadratic methods remain limited by memorization issues and a trade-off between scalability and generation quality. Our goal is to generate large graphs whose structural statistics --- e.g., degree distribution, clustering, and path length --- faithfully reflect those of real-world sparse graphs, without resorting to memorizing the training data. We introduce a discrete graph diffusion model that restricts training to a structurally motivated subset of node pairs --- observed edges and their wedge non-edges --- reducing training complexity below quadratic in the number of nodes. To keep the noisy graph informative throughout both the forward and reverse trajectories, we design a three-class absorbing forward process governed by a degree-aware, floored cosine noise schedule: unlike a structure-blind schedule that adds noise to every pair identically, ours adapts to each node's degree and never fully erases the graph's structure, keeping the noisy graph informative at every step. Experiments across diverse datasets show that our model consistently ranks among the top methods for structural fidelity against existing discrete diffusion baselines in large graph generation.

Probability-flow ordinary differential equations (PF-ODEs) are widely used as deterministic samplers for score-based diffusion models. Their usual justification is that the Fokker--Planck equation of a diffusion can be rewritten as a continuity equation driven by the score function of the forward diffusion. This identity does not, however, guarantee that the resulting velocity field generates a well-posed flow. We provide theoretical insights into the design of such deterministic samplers for generative models based on diffusions and reflected diffusions. We identify sufficient conditions for a regular Lagrangian PF-ODE flow to exist; reverse sampling and invertibility require two-sided divergence control. For learned scores, sampler stability is controlled by an unweighted velocity error, exposing a mismatch with density-weighted score matching and motivating architectural control of Jacobians, divergence, growth, and compression. Under the manifold hypothesis, positive-time regularization justifies an early-stopped PF-ODE while constants deteriorate near the data endpoint; an explicit sphere example shows that the exact deterministic flow becomes singular as the noise level vanishes even though the diffusion marginals remain well defined. These theoretical insights translate into concrete design principles for stable, invertible, and constraint-preserving diffusion samplers. We illustrate the practical relevance of these design principles using controlled numerical experiments.

Autonomous coding agents solve repository issues by reading code, running commands, editing files, and submitting patches. Extra inference-time compute yields gains only when it produces a useful repair and supplies reliable evidence for choosing one. Three behaviors decide both, and we argue they are teachable rather than byproducts of scale, so a policy can carry them instead of a scaffold. Location diversity remains narrow, since attempts return to the same site and extra samples add no coverage. Edit diversity is left unexploited, since methodologies that differ resolve complementary issues no single run reaches. Verification misleads, since a test the agent writes for its own patch accepts many incorrect ones. Directing search by execution feedback and scoring each patch against its own reverted tree resolves 52.8% of SWE-bench Verified using 48.1% of the agent-steps an eight-sample baseline spends. Training moves these behaviors into the policy. On the 270 issues held out from SFT and RL training, weighted supervised fine-tuning raises pass@1 from 31.9% to 35.2% and pass@8 from 46.7% to 51.1%. A reinforcement objective then trains the verifier against gold-labeled repairs and incorrect variants, crediting the assertions that detect them. It raises pass@1 to 43.0% and pass@8 to 60.7%, lifts verifier precision from 26.8% to 41.7%, and more than halves false acceptance. Resolution improves on two of three out-of-distribution suites and verifier precision on all three, and the gains hold at 7B, 14B, and 30B against published coder baselines.

AI agents can take actions. At times, those actions can go beyond what is intended. Agent reliability can be defined as assurance that an agent will stay within intended bounds and operate within limits. Today, there is no shared framework or language for describing, validating, and benchmarking the reliability of agentic deployments in financial services. This makes it difficult for financial institutions, vendors, and regulators to assess and trust agents at scale, thus limiting the pace of development and adoption. A standardized, shared representation of agent reliability would fill the gap. This paper introduces the Agent Reliability Profile, a per-agent unit of assurance evidence for agent deployments in financial services. Each Profile records a bounded, falsifiable claim, this agentic system reliably functions within its operating boundary. We define "operating boundary" as an agent having; (1) a defined autonomy tier, (2) a defined operational design domain, (3) defined classes of action, and (4) a defined control envelope. Production assurance progresses through three levels while the Profile schema remains constant: a Profile Builder compiles a Level 1 Asserted Profile from institutional evidence, a Profile Validator tests the deployment in its own environment to produce a Level 2 Validated Profile, and operation of the same tests by a qualified independent assessor produces a Level 3 Verified Profile. Separately a Benchmarked Profile reports results comparable across institutions under reference conditions. We describe the architecture, the artifact, the assurance ladder, the comparability flag, associated tools, an evaluation methodology, applications for financial institutions and supervisors, limitations, and a staged implementation program.

Executable Agent Skills combine natural-language instructions and scripts into reusable packages for LLM agents, and revising them requires fixing errors without breaking correct behavior. Existing benchmarks do not systematically distinguish documentation repair, script repair, and preservation when evaluating skill self-evolution. We introduce SkillScriptBench, a 350-task benchmark designed to evaluate these capabilities separately. From a survey of over 35,000 GitHub-hosted Skill roots, we select 100 packages and construct 150 repair tasks. Each task pairs a package containing injected script faults with a maintenance request and executable checks of the required behavior. A complementary controlled track contains 200 tasks from 50 packages, each evaluated under the same maintenance request in four states: clean, documentation faults, script faults, and faults in both. Across four LLMs, methods that edit both documentation and scripts can repair script faults but do not consistently outperform Markdown-only revision on documentation repair or preservation. We therefore introduce AST-Guided Skill Revision, which uses abstract syntax trees and calling relationships to link maintenance requirements to relevant code locations. It restricts script edits to these locations and updates the documentation to match the revised scripts. Averaged across models, this revision stage yields absolute gains in repair success of 21.9% for Raw Package and 27.7% for CoEvoSkills on faulty packages. Absolute gains in the proportion of tasks solved in all three runs reach 20.8% and 31.5%, respectively, indicating more consistent repair success across repeated runs.

Autonomous agents built on Large Language Models (LLMs) need standardized protocols to interoperate across systems. Several now exist (A2A, MCP, ACP, ANP, NLIP), but the Natural Language Interaction Protocol (NLIP) has not appeared in any controlled performance study, and no work has measured where an agent protocol's latency is spent. We compare NLIP and the Agent-to-Agent (A2A) protocol empirically, decomposing latency into message creation, connection, and send phases across three independent hardware environments. For lightweight coordination, NLIP is 8.4-9.6x faster than the baseline A2A SDK implementation on two environments and about 4x on a third; the direction of the advantage is consistent, its magnitude depends on the hardware. The advantage is stage-specific: for the end-to-end pipeline, where LLM inference dominates, the protocols are near parity. The difference comes almost entirely from connection setup. To test A2A at its best, we also ran A2A SDK with connection caching enabled; caching narrows its gap with NLIP by a hardware-dependent amount, from 2.75x on one machine to near-parity on faster hardware, where at scale a cache-optimized A2A-SDK matches NLIP. We report these as measured conditions without a single causal account of the residual send-phase cost. Against the more optimized Python-A2A, NLIP leads by about 4x on the same stage. We close with a protocol-selection guide keyed to workload characteristics.