post-training alignment
post-training-alignment-e386ed36·2 events·first seen 22d agoAliases: post-training alignment, alignment training
Co-occurring entities
More like this (12)
Recent events (2)
Geopolitical Bias in LLMs Originates in Post-Training, Not Pre-Training Data
A study testing seven open-weight LLM pairs (base vs. chat models) across seven labs finds that geopolitical bias is introduced during post-training rather than inherited from pre-training data. Six of seven labs showed post-training shifts favoring the developer's home country or region, with Alibaba's Qwen 2.5 showing the most extreme shift (18x increase in China-favourability log-odds). The effect is also language-dependent: Mistral becomes pro-France only under French prompting. The authors argue this implicates alignment and RLHF processes as active shapers of geopolitical perspective, calling for greater transparency and auditing of post-training pipelines.
Vision-Language Models Suppress Female Representations Under Ambiguous Input
This paper investigates gender bias in vision-language models (VLMs) when inputs are ambiguous (e.g., workers in full gear or seen from behind), finding that models default to male outputs even for strongly female-stereotyped occupations. The authors introduce LALS (Latent Association Leaning Score), a zero-shot metric that probes internal visual-token activations to measure concept associations across layers. Across 15 occupations, 800+ ambiguous images, and four VLMs, they find a systematic decoupling: models internally encode female associations but suppress them before generation, with male signals amplifying end-to-end while female signals peak mid-network and are filtered out. Cultural visual cues like clothing color further modulate these internal associations.