The paper evaluates a recurrent network that learns monotone temporal aggregation for risk scoring, comparing it to hand‑crafted monotone aggregates used in deployed pipelines. Functional regression shows the learned channel can reproduce the geometrically weighted separable family of engineered statistics with Spearman ρ ≥ 0.996, while it struggles with consecutivity, time localization, and exposure floor, all of which can be fixed with an admissible basis. In practice, the learned aggregator matches or exceeds engineered counterparts in full‑sample scenarios, offers modest AUC gains in certain decay regions, and maintains performance where tree‑ensemble scores flatten.
By Yew Lee Tan
arXiv:2606. 17572v1 Announce Type: new Abstract: Learned dynamics models often answer global physical questions, such as fault severity or impact stiffness, by pooling a per-step feature sequence into one readout vector.
By Yifan Wang
arXiv:2606. 15621v1 Announce Type: new Abstract: Per-token counterfactual credit estimation asks which token in a language-model rollout caused the final answer to be right or wrong: cut the transcript at a pivot, substitute an alternative token, replay continuations, and compare outcomes.
By Nils Matteson
arXiv:2608. 08202v1 Announce Type: new Abstract: Data-centric curation pipelines frequently rely on model confidence scores to flag and filter noisy or mislabeled training instances.
By Sai Srikar Boddupalli
arXiv:2608. 19760v1 Announce Type: cross Abstract: Audited against causal ground truth from executed replay in a single-agent tool environment (ALFWorld), none of the step-level credit signals used to train LLM agents -- LLM-judge scores, outcome-conditioned logprob ratios, or the policy's own confidence -- identifies which steps causally matter better than chance.
By Haiyue Zhang
arXiv:2608. 08126v1 Announce Type: new Abstract: Credit scoring increasingly relies on models whose decision logic cannot be read off their parameters, in tension with supervisory expectations that adverse decisions be explainable.
By Gregorius Reynaldi Pratama, Kuo-Kun Tseng
arXiv:2606. 00257v1 Announce Type: cross Abstract: Token-level credit assignment for language-model reinforcement learning is usually formulated as if the policy were fully trainable, while practical LLM-RL pipelines often rely on parameter-efficient fine-tuning, especially LoRA.
By Rodney Lafuente-Mercado
FinalityBench is an executable benchmark that tests how agents decide on shipping, re‑capturing, refunding, or waiting when a merchant’s payment processor, ledger, ERP, and bank feed receive delayed, duplicated, dropped, or reordered messages, causing contradictory beliefs about an order. The benchmark uses a hidden canonical event log and faulted delivery streams to generate system views, scoring each episode by the merchant’s terminal economic position relative to a privileged reference. It contains 321 tasks, including 45 twin pairs where all four views are identical yet the correct disposition differs, and evaluates nine programmatic policies, revealing that a ship‑on‑first‑sign policy performs best by accuracy but worst by paired loss, while a runtime‑gated irreversible‑action policy achieves 85.4% accuracy without losing money.
By Abhishek Sharma
arXiv:2607. 11022v1 Announce Type: new Abstract: The test suites used as RLVR rewards for code have natural false positives: per-task, persistent, asymmetric errors that accept the same wrong programs every time they appear, unlike the symmetric or resampled noise assumed by existing noise-robustness analyses.
By Chuyifei Zhang
arXiv:2608. 20290v1 Announce Type: new Abstract: Whether a language model has improved itself is increasingly judged not by mean accuracy but by which individual problems it gains and loses.
By Cheng Xu, Nan Yan, Liming Chen, M-Tahar Kechadi
The paper introduces a Causal Market Simulator using a continuous‑time causal prior‑data fitted network (PFN) to estimate market impact, addressing bias from informational flow that traditional models overlook. On a synthetic market prior with adjustable confounding, the PFN matches an unbiased oracle while a naive estimator overestimates by up to 68%. In zero‑shot tests on real limit‑order books and equity data, the model’s intervention responses align with ground‑truth scales and improve correlation with measured impact when a recurrent state‑space variant is used.
By Dennis Thumm, Ying Chen
Self-improving LLM agents can adapt a credit pipeline to a changed rule, but an agent that rewrites itself destroys the artefact a supervisor reviews: a named change, a recorded test, an approval. We...