Tag
论文提出 CASEL（Consequence-Aware Selective Execution Layer），一种模型无关的运行时分配策略，通过综合校准失败概率、后果严重性、可逆性、歧义性和证据新鲜度来为工具调用选择执行、验证、沙箱、澄清、审批或放弃等模式，仿真结果显示其将有害副作用轨迹比例从 14.75% 降至 7.02%。
This empirical study finds that joint upper-bound coverage of traffic speed forecasts can be substantially improved via calibration, yet improved coverage does not translate into better route-choice utility in frozen offline proxy tasks on Beijing and Chengdu road-speed data.
This paper evaluates Jev, a fixed general-purpose probabilistic decision model, for training-free human activity recognition on accelerometer data, finding it far below supervised models (macro-F1 0.038–0.118 vs 0.686–0.907) and showing that adding more numerical sensor features actually degrades performance unless paired with deterministic semantic renderings.
TabPFN-3.5 is released, claiming state-of-the-art performance for tabular data beyond IID small data settings, with features like handling grouped and temporal data, uncertainty calibration, and faster inference.
This paper proposes a risk-controlled framework for using LLMs as judges in factual evaluation, calibrating uncertainty thresholds to maintain a user-specified error rate and routing to retrieval-augmented mode when needed, achieving higher coverage with provable reliability guarantees.
The paper proposes Invertible Logits Transformation (InvLT), a post-hoc calibration method that applies a learnable scalar MLP element-wise to logits, preserving accuracy while correcting nonlinear miscalibration more flexibly than temperature scaling.
This paper introduces Unscented KalmanNet (UKN), a hybrid deep learning filter that augments the Unscented Kalman Filter with learned components to improve state estimation accuracy and covariance calibration under unknown noise statistics and model mismatch. Experiments show significant RMSE reductions over UKF and other KalmanNet variants.
Introduces an efficient Bayesian deep ensemble method for predictive regression that combines low-dimensional ensemble representation, closed-form Bayesian aggregation, and independent ensemble training to achieve calibrated uncertainty estimates with computational efficiency.
SAGE proposes a group-level uncertainty target that constructs an answer-conditioned uncertainty geometry over sampled responses to improve verbal uncertainty alignment in LLMs, and introduces GUPO for training. Experiments across reasoning tasks show improved uncertainty ranking and reduced overconfidence.
UNIQ introduces a conformal calibration method for offline reinforcement learning that adapts conservatism per-state based on uncertainty, improving over IQL on some D4RL benchmarks while maintaining memory efficiency.
UCCI proposes a calibration-first router for LLM cascades that uses isotonic regression to map token-level margin uncertainty to error probability, achieving a 31% cost reduction on a production NER workload while maintaining micro-F1=0.91 and reducing expected calibration error from 0.12 to 0.03.
This paper introduces Agent-BRACE, a method that decouples LLM agents into belief state and policy models to handle long-horizon tasks in partially observable environments. By verbalizing state uncertainty, it achieves significant performance improvements over baselines while maintaining constant context window size.
This paper introduces BitCal-TTS, a runtime controller that improves accuracy and reduces premature halting in quantized reasoning models by calibrating confidence signals during test-time scaling.