Searcharxiv⌕ Search
Home/Archive/arXiv

arXiv · 2610.23569

GCAM-Gym: A Benchmark for Agent-Based Climate Policy Search

Tianyu Zhang·Zhenghan Tai·Xinyu Wang·Irina Rish·Yoshua Bengio
View original source ↗Read source document ↗Export metadata ↓

Abstract

Comparing climate-policy search methods requires accounting for differences in policy control, available observations, feedback timing, and simulator budgets. We introduce GCAM-Gym, a Gymnasium-compatible environment that supports full-policy evaluation and stepwise control through a common interface and exposes the components of the policy objective. We compare seven search and control systems in three settings: target attainment and cost, cost-temperature trade-offs, and one-step delayed feedback. Each system is evaluated with ten seeds per setting, for a total of 3,210 attempted GCAM evaluations. Four systems share a budget of 24 attempts per seed; among these, CMA-ES achieves the highest interquartile mean (IQM) of the best returns from completed policy evaluations in all three settings. We release the evaluation archive and analysis tools to reconstruct the reported comparisons without running GCAM.

Explore related subjects

cs.CY

Keep this discovery

Explore connections, maps & timelines

BibTeXRIS

Tianyu Zhang, Zhenghan Tai, Xinyu Wang, Irina Rish, Yoshua Bengio. 2026-10-16. GCAM-Gym: A Benchmark for Agent-Based Climate Policy Search. https://arxiv.org/abs/2610.23569

Cite the original work for its findings. Save a collection to share your selection of sources.

KEEP EXPLORING

Related papers

arXiv2026

Happy Young Women, Grumpy Old Men? Emotion-Driven Demographic Biases in Synthetic Face Generation

Synthetic faces from text-to-image (T2I) models pervade digital media, yet their demographic biases under emotionally conditioned prompts remain poorly understood. We aim to systematically audit how emotionally conditioned prompts affect demographic and perceived-attractiveness biases in synthetic faces generated by T2I models, with particular attention to intersectional patterns and cross-ecosystem differences across model families. We audited eight (4 Western and 4 Chinese) T2I models and generated 56,000 faces under seven prompt conditions: a neutral baseline and six emotion conditions. We quantified biases in gender, race, age, and perceived attractiveness using information-theoretic divergence metrics. We further conducted intersectional analyses across combined demographic attributes and compared patterns between the Western and Chinese model groups to assess cross-ecosystem consistency and divergence in bias behavior. All models show strong overrepresentation of young faces, and most also overrepresent White-coded individuals. Intersectional analysis reveals compound underrepresentation or near-erasure of specific demographic combinations, such as young x female x Black faces, which are largely absent across models and are not captured by single-attribute audits. Emotion prompts act as additional demographic selectors: negatively valenced emotions (including sadness and fear) consistently shift outputs toward White, middle-aged, male-coded faces. This produces a valence-driven mapping that is also associated with lower perceived attractiveness in generated faces. These findings indicate that demographic bias in T2I face generation is both pervasive and shaped by emotional conditioning. They underscore the need for intersectional, emotion-conditioned, and multilingual demographic audits as part of standard pre-deployment evaluation practices.

cs.CY↗
arXiv2026

Animarium:Technical Report Version 1

Synthetic populations of eleven Italian municipalities (1,814,317 individuals in 887,937 households) generated from published aggregates alone: ISTAT census and register tables, census-section counts, the national civic-address register, public-use survey microdata, and six municipal open-data portals, every source certified in a registry with licence, fingerprint and declared affordances. Four rings give every attribute a declared place: a maximum-entropy joint model of up to nine demographic attributes; whole-vector donation of twenty-three attitudinal and health variables from survey respondents; placement to census section, single year of age and address; and households constrained by the census size distribution per section. Every downstream layer (detailed titles, work, names, biographies) is a declared derivation adding no information. The pipeline is deterministic to the byte: regenerating all eleven municipalities from the tagged commit reproduces every file of every ring bit for bit, in 33 minutes on one workstation. Populations are released in a public regime enforced in the data (no names, no addresses, coordinates randomised within census section), browsable in Animarium, a dependency-free web viewer where every number carries its comparison and every view is a citable URL, and downloadable as an open dataset. The report documents the architecture, the sources and their certification, the reproducibility and quality measurements at the release tag, the viewer, and the narrative layer that renders records into personas for LLM-driven simulation, with the platform's controllability demonstrated in companion experiments, and validation explicitly out of scope.

cs.CY↗
arXiv2026

Can AI interview? Comparing AI-led to Human-led Chat Interviews

Large language model chatbots have been proposed as AI-interviewers for collecting qualitative interview data at scale, potentially bridging social science's longstanding divide between the breadth of surveys and the depth of qualitative interviews. This paper provides the first direct empirical comparison of data quality in AI-led and human-led interviews. In a randomized experiment (N = 41), university students were interviewed via synchronous text chat by either trained non-expert interviewers or a locally hosted AI system; we evaluate exposure, specificity, and relevance. Median interview duration was similar across conditions, but AI-interviewers asked more than twice as many questions per interview and elicited roughly 65 percent more respondent text per interview. Per-question responses were longer with human interviewers, though not after multiple-testing correction; only AI-condition participants skipped questions. Responses were almost uniformly on-topic in both conditions; for specificity we detect no difference, but the design cannot exclude one of up to three quarters of a point. These results were obtained with a small, locally hosted open-weight model (7B parameters), which addresses concerns about data security, GDPR compliance, and scientific reproducibility. Our findings support the viability of AI-interviewing for large-N qualitative data collection.

cs.CY↗