Science & Mathematics

Game theory identifies when a strategic action should be accepted, but not how long a live, language-based counterpart will continue to accept it. We entered the GLEE competition in its final week with one protected agent, Milton, and four instrument agents that ran deterministic, per-game randomized trials of policy changes; Milton received changes only through pre-specified gates or registered, rollback-guarded production rules. In negotiation, learned pricing improves adjusted percentile by 0.071 (n=21,168); in bargaining, a continuation model conditioned on the rejected offer did not improve on a recency-weighted incumbent and was retired. In persuasion, a preregistered five-dose experiment (7,391 assigned games, blinded until a fixed stop) shows that recommending every unit raises sales by 0.194 of the maximum (95% CI [0.147, 0.241]) where the receiver's slack makes obedience rational (52% of the Bayesian benchmark), lowers them where it does not (-0.067), and produces a larger early-to-late purchase decline than honest recommendations (-0.123). A rule fixed before unblinding now routes Milton's dose per cell. Milton stood 11th of 460 agents in bargaining at the fixed stop; we also report the close, where the final day's pool shift moved all five agents' bargaining ratings together while their play was unchanged, and three dispatch defects were found by checking delivered behavior against the design.

Reliable anomaly detection is a deployment bottleneck for 5G/Open RAN observability and emerging self-optimizing 6G systems, where hundreds of heterogeneous KPIs, severe class imbalance, and strict alert budgets make purely numerical scores brittle. We present KAC, a deployment-oriented KPI-aware multimodal anomaly-detection framework that uses KPI-level residual-text alignment to turn compact operator-style summaries into a learnable feedback signal over full-KPI temporal evidence. This gives the detector broad numerical coverage like modern time-series models, while retaining the sparse KPI-specific reasoning that makes operator feedback sustainable in a future operator-facing workflow. Across public telecom telemetry, enterprise-scale Open RAN observability data, and a Nokia benchmark built from real 5G RAN production packet traces with realism-hardened injected anomalies, KAC achieves strong threshold-based detection performance while remaining competitive on threshold-free ranking metrics. On TelecomTS, at a fixed 2.5 percent false-positive-rate operating point, KAC achieves the highest recall among the evaluated residual-based methods while maintaining comparable precision; across the evaluated benchmarks, it remains robust under substantial class imbalance. KAC is integrated in pre-production shadow mode with Nokia's GKE-based KPI construction pipeline, where it continuously scores telemetry windows alongside an internal packet-level anomaly detector for operational evaluation.

Model-based reinforcement learning (MBRL) has enabled sample-efficient learning of continuous control policies from high-dimensional observations through learned latent dynamics models. These models are typically trained with observation reconstruction objectives, which provide strong supervision but can bias representations toward visual fidelity leading to an undertrained prior and unreliable long-horizon predictions. We propose a stochastic gating mechanism that induces controlled partial observability during training by sampling a spatial binary mask from a Beta--Bernoulli random field and applying it to the raw observation. The gate probability is resampled at each time step, exposing the encoder to conditions ranging from near-total occlusion to near-full visibility. Integration of our method requires minimal architectural changes from existing dynamics models: the gated observation and the binary mask are channel-wise concatenated and passed to the encoder as a single augmented input, with no modifications to the broader training pipeline. We evaluate the proposed approach on a set of pixel-based continuous control tasks selected from the DeepMind Control Suite. Using model predictive control for evaluation, which relies solely on open-loop model rollouts, we show that stochastic gating improves planning performance and data efficiency across tasks. Further analysis demonstrates that performance is robust to the spatial scale of the mask, and a balanced observability during training outperforms configurations biased toward either sparser or denser observations.

AI assistants may carry forward not only observations but also interpretations of what a person is doing or may do next across time. Once retained, these interpretations become context for later predictions and system-triggered outputs. We study whether this representation step changes later behavior in gaze-aware streaming assistance. Holding retrieved events, their order, and the target input fixed, we find that Free-form Interpretation amplifies historical influence in both directions: correct histories produce more helpful changes and wrong histories more harmful changes. Unsupported and broader interpretations about activity or future behavior are associated with greater harmful influence. Evidence-Bounded Interpretation preserves distinctions between observation, inference, uncertainty, and temporal scope, reducing unsupported and overcertain content and lowering wrong-source harmful influence while retaining most helpful influence. The same effect structure appears with a second model. These findings identify persistent interpretation as a consequential design choice for history-aware human-AI systems.

Agents that work for many turns outgrow the context window of the model that serves them, and on small language models (SLMs) and constrained inference systems that window is small. Context compaction replaces older turns with a summary so a session can continue, but it is lossy, costs a model call, and can help or hurt depending on when it runs. We describe Bud ContextCompact 2.0, the compaction built into the Bud agent runtime, and evaluate it in two settings. On recall questions over recorded agent sessions, compaction to a 16K budget raised a 1.2B-parameter model's accuracy to 0.335, from 0.247 with truncation and 0.176 with its full 32K history (p < 0.001), and raised a 9B model from 0.515 (truncation) to 0.577 (p = 0.04). On 100 SWE-bench Verified tasks run end to end by a 27B reasoning model on a 49K-token window (556 graded sessions), compaction raised the resolve rate on long sessions from 41% to 53% (+12.3 points, 95% CI +1.6 to +23.0, McNemar p = 0.043), left short sessions unchanged, and used 13% fewer tokens per solved task at equal total cost. The main failure without compaction was not the history overflowing but a reasoning turn exhausting the room the history left. The gain has limits: it vanished under a 45-minute session budget, because each summary took a median of 161 s, and in a threshold study it appeared only when compaction triggered early enough to leave room. An 8B model submitted after about ten turns at 5-8K tokens and never needed compaction. We argue that the accuracy results transfer across hardware, including CPU inference, for an equivalent model and window, while the time cost scales with the hardware and must be budgeted for.

We introduce ReDT, a feed-forward visual geometry transformer designed to scale multi-view reconstruction to high-definition (HD) inputs. Existing visual geometry transformers are typically trained at fixed, standard definition (SD) resolutions and exhibit degraded reconstruction quality and poor computational scaling to high input resolutions. While recent efficient architectures reduce the complexity of inter-frame attention, expensive intra-frame and encoder attention remain largely unaddressed, making HD reconstruction impractical with existing architectures. ReDT addresses both limitations with a resolution-aware architecture designed to scale to high-resolution inputs. The key idea of ReDT is to first map each view to a fixed-size set of basis tokens, compute correspondences in this compact constant-size basis space, and then project the outputs back to the original resolution for dense prediction. By decoupling the internal representation from input resolution during attention, ReDT substantially reduces the computational cost of high-resolution 3D reconstruction and improves model performance at all resolutions. Extensive experiments on multiple datasets spanning SD to FHD (2K) show that ReDT outperforms state-of-the-art visual geometry transformers in both reconstruction quality and computational efficiency. For instance, on the ETH3D (2K) dataset, ReDT achieves a $16\times$ speedup, a $34\%$ reduction in peak GPU memory usage, and a $33\%$ reduction in chamfer distance over VGGT-$Ω$. \href{https://redt.github.io}{Project Website.}

We show that two-layer transformers trained to near-zero loss on a sorting task converge to one of two mechanistically distinct solution classes depending on their initialization. One solution class consists of solutions that do not use the second attention layer to sort, but also fail to length-generalize. The other, which uses both attention layers, length-generalizes perfectly, despite only being trained to sort sequences of a fixed length. By carefully choosing the weight initialization, we demonstrate that we can steer the model toward learning one solution or the other. We identify input types that are rare at the training length but increasingly common at longer sequence lengths as a source of this disparity in length-generalization, and find that layer-wise specialization of the second class of solutions allows them to handle these rare sequences more accurately. In the second class, we discover a mechanism by which the second layer attention can use the contextual information gathered by the first layer attention to reduce its search space. On the related task of finding the minimum of a sequence, we find a similar two solution-type phenomenon, but with the opposite generalization pattern, where a different form of layer-wise specialization causes the two-layer solution to length-generalize worse than the single-attention solution.

Test-time scaling (TTS) improves large language model (LLM) reasoning by sampling multiple answers, but fixed sampling budgets ignore both query difficulty and the changing energy cost of inference under GPU contention. We measure up to a 4.1-fold increase in energy for the same four-sample query as GPU load rises. We present ATEST, a resource-aware TTS framework that jointly considers the expected value of additional samples and their current GPU cost. ATEST-Online, its default mode, requires no offline calibration: it incrementally generates samples, estimates the stability of the plurality answer, monitors live GPU conditions, and learns step energy cost online to decide when to stop. ATEST-Calibrated instead selects a per-query budget using offline accuracy and cost models, favoring deployments with stable loads. We evaluate both modes on four reasoning and question-answering datasets using Phi-3-mini on an NVIDIA A100 under controlled varying load, and test transfer to an NVIDIA L40S and Qwen2.5-3B. Across the four A100 datasets, ATEST-Online answers every query without exceeding the energy budget, while achieving the highest or statistically tied accuracy and reducing sampling as load increases. On GSM8K at high load, it achieves 60.0% accuracy with no deferrals, versus 44.0% accuracy and 42.4% deferrals for ATEST-Calibrated. Ablations show that online cost learning drives adaptation to load. Including offline profiling, ATEST-Online uses 42 times less GPU time and 48 times less energy than ATEST-Calibrated for the evaluated workload. Online control trades higher tail latency and lower throughput on short tasks for robustness under varying load; offline calibration remains competitive under stable conditions.

Conversational large language models can respond with warmth and reassurance while failing to advance the user's stated goal. We conceptualize this failure as $\textbf{goal displacement}$, the substitution of persona-maintenance content for goal-relevant substance. To measure it, we introduce $\textbf{GDB}$ (Goal Displacement Benchmark), comprising 230 scenarios across nine domains and $4{,}134$ human-scored responses from 18 models. A $2 \times 2$ taxonomy crosses hard constraints with emotional distress, distinguishing failures to confront infeasible goals or combine empathy with practical guidance from reverse displacement, where unsolicited problem-solving replaces requested emotional support. A deletion-test rubric assesses whether substantive help remains after purely supportive sentences are removed. Independent raters achieve weighted $κ= 0.82$ against author scores for displacement severity. Trap displacement ranges from $0.0\%$ to $37.7\%$ and is strongly negatively correlated with Chatbot Arena Elo ($ρ= -0.706$, $n = 15$). Reverse displacement reaches $41.9\%$, including substantial rates in models with negligible trap displacement. A factorial probe supports a distinction from social sycophancy. Human-scored responses in the high-complexity, low-pressure condition exhibit $35.0\%$ sycophancy but $7.5\%$ displacement, with little association between the constructs. Matched OLMo-2 checkpoints show that mixed empathy-with-constraints displacement remains at $52.3\%$ after supervised fine-tuning and falls to $13.8\%$ after direct preference optimization. GDB provides a reproducible diagnostic of whether responses advance the operationalized user goal, supporting model comparison, post-training evaluation, and regression testing. We release the dataset, annotations, and rubric under CC-BY-4.0 and the evaluation code under MIT.

Time series are ubiquitous across different domains, and the growing use of synthetic data for privacy-preserving sharing, data augmentation, and stress testing makes time series generation increasingly important. Early generative approaches are predominantly autoregressive, which can suffer from low sample diversity, error accumulation, and difficulty modeling long-range dependencies. In this paper, we introduce FracTS, a hierarchical, coarse-to-fine autoregressive model for multivariate time series generation. Inspired by fractal generative modeling developed for images, FracTS adapts the paradigm from 2D to 1D sequences by learning a multi-scale representation that captures global trajectory structure at coarse scales and local dynamics at fine scales while following the autoregressive nature of time series. In contrast to unstructured images, time series are structured, allowing easy automated or designated temporal feature processing optimized for the model, and global (static) feature extraction describing the instances that generate the corresponding time series in the collection. We incorporate them in the generation process to further improve the generation quality. The design naturally handles multivariate dependencies and improves long-horizon temporal coherence without requiring fixed-length inputs. Experiments on five datasets (two conditional, three unconditional) conducted against seven baselines show that FracTS consistently outperforms baselines across marginal distribution and predictive ML utility, and correlation preservation in the unconditional setting. In particular, the predictive ML utility is averagely improved by 100.0% and 75.0-81.0% for unconditional and conditional time series datasets compared to the best baseline (relative error computed on dataset-averaged raw values). Code is available at: https://github.com/BetterdataLabs/fracts.

The increasing connectivity of adaptive cruise control (ACC) systems exposes vehicle-state information to cyberattacks, potentially causing irrational car-following behavior and unsafe longitudinal dynamics. Existing trajectory recovery methods generally lack an interpretable criterion for recovery activation and remain vulnerable to attack-induced distribution shifts. To address these challenges, this paper proposes TRACER, a rational driving constraint (RDC)-triggered trajectory recovery framework for cyberattack-resilient ACC systems. TRACER employs a rationality-aware triggering (RAT) mechanism that activates recovery when the observed driving state violates the calibrated RDC bounds. It also introduces cross-branch robust imitation learning (CRIL), which combines a clean branch for learning nominal driving behavior with a noise-enhanced branch for improving robustness to corrupted observations. Cross-branch fine-tuning further aligns the robust recovery policy with the nominal driving-state distribution. Experiments across six cyberattack scenarios show that TRACER achieves reliable recovery triggering and outperforms state-of-the-art recovery baselines in terms of accuracy, stability, and safety. State-space analysis further confirms that TRACER mitigates attack-induced distribution shifts and restores coupled spacing-velocity dynamics, highlighting the effectiveness of robust policy learning for cyberattack-resilient ACC.

We give the first unconditional family of polynomial-time algorithms that breaks the $1/3$ approximation barrier for symmetric bimatrix games. For every fixed $δ>0$, an algorithm in this family computes a symmetric $(1-1/\sqrt{2}+δ)$-approximate Nash equilibrium, improving on the previous best guarantee of $(1/3+δ)$ due to Kontogiannis and Spirakis [2011].

Coding agents keep working on runs that will fail verification, and asking a frontier model to watch for this is expensive in its own right. We introduce StopLoss, a small external monitor that learns failure recognition from completed trajectories and their verifier outcomes. The name is inspired by the stop-loss order in finance: StopLoss ends a run early when the run is likely to fail, which saves resources. Final outcomes label earlier local windows, so a fine-tuned language model can score the most recent tool steps in a single forward pass, and a threshold calibrated on successful runs turns that score into an abort decision without changing the executor. Trained on GPT-family SWE-rebench trajectories and evaluated on SWE-bench Verified and a SWE-bench Pro subset with GPT and Claude executors, StopLoss is compared against six general-purpose judges reading identical windows. Outcome supervision lifts failure prediction from near chance to a useful signal at the first check, at one to two orders of magnitude lower cost per check than any tested judge. This translates into savings: relative to completing every task, StopLoss lowers cost per retained solve by 4.5\% on Verified while preserving 99\% of successful runs, and by up to 31\% on the costlier Pro subset, reaching 48\% with a Qwen3-8B backbone, and in all four settings its cost per retained solve is below that of every judge policy. Outcome-trained small monitors can make failure recognition cheap enough for early stopping to pay for itself.

We analyze the reconstruction error of a low-dimensional Hermite--Gaussian (HG) ptychographic method used in real-time Wigner distribution deconvolution (WDD) reconstruction. We derive an upper bound on the discrepancy between the low-dimensional, aberration-corrected reconstruction and classical Fourier-domain WDD. The bound decomposes the reconstruction error into contributions from HG--Fourier basis mismatch and aberration-coefficient mismatch, explicitly characterizing their dependence on the number of retained HG modes, Wiener regularization, aberration mismatch, and measurement dose. Numerical experiments on simulated electron and X-ray ptychographic data characterize these dependencies and demonstrate robustness to Poisson noise. These results provide theoretical and guidance for selecting the reduced representation and reconstruction parameters, enabling computationally and memory-efficient for near-real-time ptychographic reconstruction.

Determining OpenCL channel depths for Decoupled Access-Execute (DAE) architectures in High-Level Synthesis (HLS) is critical for balancing performance and hardware efficiency. Conventional static analysis tools often over-provision these memory channels to ensure performance safety, leading to significant resource stranding. This paper presents an expansion of an LLVM-based static analysis framework, originally proposed for automated DAE generation, by introducing a hardware-aware refinement layer. We identify a counter-intuitive Performance Inversion Paradox linked to baseline memory stalls, where high-stall, grid-based kernels reach peak performance at minimal channel depths and suffer up to a 61% throughput degradation when depths are increased. We propose a Heuristic Refinement Layer that optimizes OpenCL channel depths based on this correlation between baseline stalls and performance saturation. Experimental results demonstrate that our expanded framework successfully identifies the Minimal Peak Performance point across various benchmarks. By avoiding the performance traps of standard over-provisioning, we achieve a 2.52x improvement in Resource Efficiency for an illustrative HotSpot application. This approach maintains peak execution speedup while ensuring the leanest possible hardware profile, outperforming conservative tool-based static predictions.