Lowest Rated Papers

Rating: Lowest Clear all

Advanced filtersTopic: All

Experimental

Topic

Topics are auto-detected from title, abstract, and metadata and may be imperfect.

Publication Date

Newest Oldest

Average Rating

Highest Lowest Clear

Large Causal Models from Large Language Models

Sridhar Mahadevan

arXiv·2025

We introduce a new paradigm for building large causal models (LCMs) that exploits the enormous potential latent in today's large language models (LLMs). We describe our ongoing experiments with an implemented system called DEMOCRITUS (Decentralized Extraction of Manifold Ontologies of Causal Relations Integrating Topos Universal Slices) aimed at building, organizing, and visualizing LCMs that span disparate domains extracted from carefully targeted textual queries to LLMs. DEMOCRITUS is methodologically distinct from traditional narrow domain and hypothesis centered causal inference that builds causal models from experiments that produce numerical data. A high-quality LLM is used to propose topics, generate causal questions, and extract plausible causal statements from a diverse range of domains. The technical challenge is then to take these isolated, fragmented, potentially ambiguous and possibly conflicting causal claims, and weave them into a coherent whole, converting them into relational causal triples and embedding them into a LCM. Addressing this technical challenge required inventing new categorical machine learning methods, which we can only briefly summarize in this paper, as it is focused more on the systems side of building DEMOCRITUS. We describe the implementation pipeline for DEMOCRITUS comprising of six modules, examine its computational cost profile to determine where the current bottlenecks in scaling the system to larger models. We describe the results of using DEMOCRITUS over a wide range of domains, spanning archaeology, biology, climate change, economics, medicine and technology. We discuss the limitations of the current DEMOCRITUS system, and outline directions for extending its capabilities.

No ratings yet

View paper →

Association of the 2018 Japanese Financial Incentive Policy With Long-Term Care Needs and Mortality

Daisuke Kato, Ichiro Kawachi, Tami Saito, Yuichi Yasufuku, Naoki Kondo

Asia Pacific Journal of Public Health·2025

Japan launched a municipal financial incentive program in 2018 to promote higher-quality long-term care (LTC). We quantified the association between municipal incentive performance and subsequent incident LTC certification and all-cause mortality among community-dwelling older people. Baseline 2016 data from the Japan Gerontological Evaluation Study (N = 90 896 adults aged ≥65 years in 23 municipalities) were linked to 2018-2020 administrative records. Each municipality received a composite incentive score (0-612). Using multivariable logistic regression with fixed effects, we estimated odds ratios (ORs) per 1 SD score increase, adjusting for demographics, socioeconomic status, and health. Over three years, 10.7% newly required LTC certification and 5.1% died. A 1 SD higher incentive score was associated with lower odds of incident LTC (OR: 0.91, 95% CI [0.90, 0.92]) and all-cause mortality (OR: 0.66, 95% CI [0.54, 0.81]). These benefits were driven by the “Support Functional Independence/Prevent Progression” domain in less-urban areas, while plan-do-check-act–style activities predominated in cities. Overall, better municipal performance under Japan’s financial incentive policy correlated with short-term reductions in LTC utilization and mortality among older residents, supporting continued monitoring and policy refinement.

No ratings yet

View paper →

PreviousPage 363 of 608Next

Towards eliciting latent knowledge from LLMs with mechanistic interpretability

Bartosz Cywiński, Emil Ryd, Senthooran Rajamanoharan, Neel Nanda

arXiv·2025

As language models become more powerful and sophisticated, it is crucial that they remain trustworthy and reliable. There is concerning preliminary evidence that models may attempt to deceive or keep secrets from their operators. To explore the ability of current techniques to elicit such hidden knowledge, we train a Taboo model: a language model that describes a specific secret word without explicitly stating it. Importantly, the secret word is not presented to the model in its training data or prompt. We then investigate methods to uncover this secret. First, we evaluate non-interpretability (black-box) approaches. Subsequently, we develop largely automated strategies based on mechanistic interpretability techniques, including logit lens and sparse autoencoders. Evaluation shows that both approaches are effective in eliciting the secret word in our proof-of-concept setting. Our findings highlight the promise of these approaches for eliciting hidden knowledge and suggest several promising avenues for future work, including testing and refining these methods on more complex model organisms. This work aims to be a step towards addressing the crucial problem of eliciting secret knowledge from language models, thereby contributing to their safe and reliable deployment.

No ratings yet

View paper →

Reasoning Models Don’t Always Say What They Think

Yanda Chen

Anthropic

Chain-of-thought (CoT) offers a potential boon for AI safety as it allows monitoring a model’s CoT to try to understand its intentions and reasoning processes. However, the effectiveness of such monitoring hinges on CoTs faithfully representing mod- els’ actual reasoning processes. We evaluate CoT faithfulness of state-of-the-art reasoning models across 6 reasoning hints presented in the prompts and find: (1) for most settings and models tested, CoTs reveal their usage of hints in at least 1% of examples where they use the hint, but the reveal rate is often below 20%, (2) outcome-based reinforcement learning initially improves faithfulness but plateaus without saturating, and (3) when reinforcement learning increases how frequently hints are used (reward hacking), the propensity to verbalize them does not increase, even without training against a CoT monitor. These results suggest that CoT mon- itoring is a promising way of noticing undesired behaviors during training and evaluations, but that it is not sufficient to rule them out. They also suggest that in settings like ours where CoT reasoning is not necessary, test-time monitoring of CoTs is unlikely to reliably catch rare and catastrophic unexpected behaviors.

No ratings yet

View paper →

Railroads and American Economic Growth: A “Market Access” Approach

Dave Donaldson, Richard Hornbeck

This article examines the historical impact of railroads on the U.S. economy, with a focus on quantifying the aggregate impact on the agricultural sector in 1890. Expansion of the railroad network may have affected all counties directly or indirectly—an econometric challenge that arises in many empirical settings. However, the total impact on each county is captured by changes in that county’s ‘‘market access,’’ a reduced-form expression derived from general equilibrium trade theory. We measure counties’ market access by constructing a network database of railroads and waterways and calculating lowest-cost county-tocounty freight routes. We estimate that county agricultural land values increased substantially with increases in county market access, as the railroad network expanded from 1870 to 1890. Removing all railroads in 1890 is estimated to decrease the total value of U.S. agricultural land by 60%, with limited potential for mitigating these losses through feasible extensions to the canal network or improvements to country roads

No ratings yet

View paper →

Hindsight Experience Replay

Marcin Andrychowicz, Filip Wolski, Alex Ray, Jonas Schneider, Rachel Fong, Peter Welinder, Bob McGrew, Josh Tobin, Pieter Abbeel, Wojciech Zaremba

arXiv·2017

Dealing with sparse rewards is one of the biggest challenges in Reinforcement Learning (RL). We present a novel technique called Hindsight Experience Replay which allows sample-efficient learning from rewards which are sparse and binary and therefore avoid the need for complicated reward engineering. It can be combined with an arbitrary off-policy RL algorithm and may be seen as a form of implicit curriculum. We demonstrate our approach on the task of manipulating objects with a robotic arm. In particular, we run experiments on three different tasks: pushing, sliding, and pick-and-place, in each case using only binary rewards indicating whether or not the task is completed. Our ablation studies show that Hindsight Experience Replay is a crucial ingredient which makes training possible in these challenging environments. We show that our policies trained on a physics simulation can be deployed on a physical robot and successfully complete the task.

No ratings yet

View paper →

Lowest Rated Papers

Large Causal Models from Large Language Models

Association of the 2018 Japanese Financial Incentive Policy With Long-Term Care Needs and Mortality

Google Search

Why Not 100% Equities

Towards eliciting latent knowledge from LLMs with mechanistic interpretability

Unadjusted Lower Limit of Normal for Airflow Obstruction

Reasoning Models Don’t Always Say What They Think

Doing Without — The Rupture

Living a piety-led life beyond Muharram: becoming or being a South Asian Shia Muslim in the UK

W19213

Railroads and American Economic Growth: A “Market Access” Approach

Hindsight Experience Replay