
Large Language Models (LLMs) have made remarkable breakthroughs in reasoning, yet continue to struggle with hallucinations, logical errors, and inability to self-correct during complex multi-step tasks. Current approaches like chain-of-thought prompting offer limited reasoning capabilities that fail when precise step validation is required. We propose Environment Augmented Generation (EAG), a framework that enhances LLM reasoning through: (1) real-time environmental feedback validating each reasoning step, (2) dynamic branch exploration for investigating alternative solution paths when faced with errors, and (3) experience-based learning from successful reasoning trajectories. Unlike existing methods, EAG enables deliberate backtracking and strategic replanning through tight integration of execution feedback with branching exploration. Our a1-32B model achieves state-of-the-art performance among similar-sized models across all benchmarks, matching larger models like o1 on competition mathematics while outperforming comparable models by up to 24.4 percentage points. Analysis reveals EAG's distinctive scaling pattern: initial token investment in environment interaction yields substantial long-term performance dividends, with advantages amplifying proportionally to task complexity. EAG's theoretical framework demonstrates how environment interactivity and systematic branch exploration together establish a new paradigm for reliable machine reasoning, particularly for problems requiring precise multi-step calculation and logical verification.
Long contexts break transformers: attention scores dilute across thousands of tokens, critical information gets lost in the middle, and the model cannot adapt to novel patterns at inference time. Recent work on test-time adaptation addresses this by maintaining a form of working memory—transient parameters updated on the current context—but existing approaches employ uniform write policies that waste computation on low-value regions and suffer from high gradient variance across semantically heterogeneous contexts. In this work, we reframe test-time adaptation as a budget-constrained memory consolidation problem, asking: given limited computational budget, which parts of the context should be consolidated into working memory? We propose GDWM ( G ated D ifferentiable W orking M emory ), a framework that introduces a Write Controller to gate the memory consolidation process. Our controller estimates Contextual Utility—an information-theoretic measure quantifying how much each region depends on long-range context—and allocates gradient steps accordingly, subject to a coverage constraint that ensures global representation. Theoretically, we prove that our chunk-restricted sampling strategy reduces gradient variance by eliminating inter-chunk variance via the Law of Total Variance. Experiments on ZeroSCROLLS and LongBench v2 benchmarks demonstrate that GDWM achieves comparable or superior performance with 4 ×fewer gradient steps compared to uniform baselines—excelling on sparse-information tasks (+6–13% on Qasper, +5–13% on GovReport for smaller models) while revealing principled trade-offs on dense-coverage tasks, establishing a new efficiency-performance Pareto frontier for test-time adaptation.
We propose a novel machine learning approach for forecasting the distribution of stock returns using a rich set of firm-level and market predictors. Our method combines a two-stage quantile neural network with spline interpolation to construct smooth, flexible cumulative distribution functions without relying on restrictive parametric assumptions. This allows for accurate modelling of non-Gaussian features such as fat tails and asymmetries. Furthermore, we show how to derive other statistics from the forecasted return distribution, such as the mean, variance, skewness, and kurtosis. The derived mean and variance forecasts offer significantly improved out-of-sample performance compared to standard models. We demonstrate the robustness of the method in U.S. and international markets. (c) 2026 International Institute of Forecasters. Published by Elsevier B.V. All rights are reserved, including those for text and data mining, AI training, and similar technologies.
Large language models produce fluent multi-document summaries, but their attributions are typically coarse—whole documents or passages—and generated post hoc, leaving each statement hard to verify. We argue that attribution should be a structural property of generation rather than a downstream prediction. We present CAMS, a Claim-Anchored Multi-document Summarization framework that decomposes every source document into atomic claims whose provenance is resolved deterministically from verbatim quotes to token spans, clusters equivalent claims across documents while flagging inter-source conflicts, selects a support-aware and salient subset, and rewrites it so that every summary sentence terminates in claim identifiers resolving back to source spans. This yields a separation we make explicit: provenance is an invariant holding for every emitted sentence independently of model accuracy, whereas faithfulness is an objective that selection, constrained rewriting, and verification only encourage—a distinction end-to-end and post-hoc systems conflate. We evaluate on MultiNews, DiverseSumm, and zero-shot on WCEP under a two-regime protocol separating reference-free citation quality from gold-aligned localization, audited by a support model never used for selection or verification. CAMSmatches strong end-to-end and span-attribution baselines on summary quality while improving faithfulness and citation precision, raising multi-source attribution accuracy from 38
Abstract The emergence of generative artificial intelligence (GenAI) has prompted universities to develop policies regulating its use in teaching, learning, and assessment. Little empirical research has examined how they conceptualise assessment, and the assumptions underpinning their normative frameworks. This study addresses this gap through a comparative analysis of GenAI-related policy documents from 14 universities in Switzerland and Spain. Drawing on Chan's AI Ecological Education Policy Framework and Bacchi's problematisation theory, we use collaborative consensus-based coding to examine pedagogical, governance, and operational dimensions alongside problem representations, assumptions, subject positions, and silences. The findings reveal coexisting and sometimes competing representations of GenAI. Concerns about academic integrity, authorship, transparency, and data protection coexist with calls for AI literacy, assessment redesign, formative feedback, and critical engagement with AI. Assessment emerges as a key site of this tension, with policies combining AI-resistant approaches with attempts to integrate GenAI into educational processes. The analysis also reveals asymmetrical subject positions: teachers and institutions generally define legitimate AI use, whereas students are positioned as accountable users expected to disclose, justify, or verify their use of AI. AI literacy and evaluative judgement are likewise framed mainly as students competencies, leaving teachers corresponding needs less examined. Current GenAI policy therefore represent an important but incomplete response to AI-driven change. Future policy development should move beyond regulating acceptable use towards participatory and pedagogically grounded frameworks that involve students more actively in shaping AI-mediated assessment and support human agency, evaluative judgement, and meaningful learning.