TLDRocket
Sign in

AI Causality & Inference

2 summarised stories about AI Causality & Inference, each linking back to the original source. Browse all topics →

Thursday, 16 July 2026

Scaling Point-in-Time Language Models

arXiv cs.CL 18 hours ago

Researchers developed point-in-time language models trained only on text available up to specific calendar dates to eliminate lookahead bias for financial and social science applications. Models with up to 4 billion parameters trained on 1 trillion chronologically filtered tokens achieved performance approaching comparable unrestricted models like Gemma-3-4B and LLaMA-7B across common reasoning and language understanding benchmarks. The release of the training pipeline and dataset enables reproducible temporal validity research that was previously compromised by models inadvertently using future information.

Partially Observed Structural Causal Models

arXiv cs.AI 18 hours ago

Researchers introduced Partially Observed Structural Causal Models (POSCMs), extending structural causal models to handle settings where hidden contexts simultaneously determine both the interaction structure and downstream mechanisms on observed variables. The framework was validated using two simulators: a biophysically detailed virtual human retina and a gene-regulatory system, with experiments demonstrating identifiability conditions under various latent variable scenarios. This enables causal inference in complex systems where graph structure and mechanisms are jointly generated and partially unobserved.