4arXiv cs.CL (Computation and Language)·
Study finds RAG transfers ideological bias into LLM outputs, with temperature modulating strength of transfer
A new arXiv paper investigates how ideological content in retrieved documents propagates into LLM-generated answers under a RAG framework. Using Lexical Multidimensional Analysis on 1,117 COVID-19 treatment articles as the retrieval corpus, the authors find that RAG reliably transmits ideological discourse into model outputs, with the effect peaking at moderate sampling temperatures and diminishing at low temperatures where deterministic sampling suppresses discourse transfer. The work highlights an underexplored bias vector in RAG pipelines beyond factual hallucination.
Related guides (2)
Related events (8)
AI-Mediated Communication Can Steer Collective Opinion via LLM Editing Biases
This paper demonstrates empirically that LLMs from multiple model families introduce directional biases when editing human-written texts on contested topics (e.g., nudging toward gun control, against atheism). The authors develop a mathematical opinion-dynamics model showing these biases are amplified through social networks, shifting collective opinion at scale. An audit of X's 'Explain this post' feature finds evidence of pro-life bias in Grok's outputs on abortion content, traced to specific design choices. The paper concludes with implications for EU legislative efforts on AI-mediated communication.
Finetuning on narrow datasets causes broad ideological shifts in LLMs, including extremist outputs
A new arXiv paper demonstrates that finetuning LLMs on small, moderation-passing datasets with ideological slant causes broad ideological shifts across unrelated domains — a phenomenon the authors call 'ideological generalisation.' Training GPT-4.1 on economics Q&A with a political lean shifts model outputs on criminal justice, environment, and cultural topics, and can produce out-of-distribution extremist endorsements (e.g., race-IQ connections, political violence) not present in training data. The effect replicates on Gemma-3, survives mixing with generic data, and preserves general capabilities (GSM8K within ±1pp). The finding has significant implications for supply-chain safety of finetuned models deployed via third parties.
LLMs fail to consistently simulate demographic perspective-taking in hate speech annotation
A new arXiv paper evaluates whether persona-conditioned LLMs can replicate how different demographic groups perceive hate speech, testing three dimensions: inter-group disagreement, in-group sensitivity, and vicarious prediction. No model consistently captures all three dimensions, and performance is highly model-dependent rather than emerging reliably from identity prompts alone. Vicarious prompting with Llama 3.1 provides the closest approximation to human disagreement patterns across demographic axes. The findings have implications for using LLMs as proxies for diverse human annotators in content moderation tasks.
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.
Political Consistency Training: Reducing Covert Political Bias in LLMs via RL
Researchers identify a phenomenon called 'covert political bias' in LLMs, where models handle politically paired topics asymmetrically across 7 identified technique categories. They propose two metrics—Sentiment Consistency and Helpfulness Consistency—to measure this asymmetry. To address it, they introduce Political Consistency Training (PCT), an RL-based method with complementary training paradigms that reduces covert bias while preserving overall helpfulness and generalizing to held-out benchmarks.
Contagion Networks: formal framework for measuring evaluator bias propagation in multi-agent LLM systems
A new arXiv preprint introduces Contagion Networks, a formal framework for quantifying how systematic evaluation biases spread across interacting LLM agents in multi-agent systems. Using a controlled 3-agent experiment with DeepSeek-chat, the authors measure a Cross-Agent Contagion Matrix and find that homogeneous-model agents produce contagion coefficients 3-5x weaker than cross-model settings. A key practical finding is that increasing evaluator committee size from k=1 to k=3 reduces effective contagion by 72.4%, offering a concrete mitigation strategy. The authors release an open-source experimental framework alongside the paper.
Multi-agent semantic rewriting framework for privacy-preserving RAG
A new arXiv preprint proposes a three-agent framework for sanitizing retrieved content in RAG pipelines by performing privacy extraction, semantic analysis, and reconstruction as an offline preprocessing step. Evaluated on ChatDoctor and Wiki-PII datasets across six LLMs, the approach reduces targeted information exposure in LLaMA-3-8B from 144 baseline instances to 1, while maintaining contextual fidelity (BLEU-1 of 0.122 vs. SAGE's 0.117). The framework introduces no additional online inference latency since rewriting is done offline. Source code is publicly released.
Study finds LLM-generated research ideas cluster around synthesis and bridging, diverging from human distribution
A new arXiv paper introduces a large-scale evaluation framework for comparing LLM-generated research ideas against human-authored ones, using reverse-engineered prior-work sets as prompts. The authors develop a two-axis taxonomy of research taste (opportunity pattern and research paradigm) and find a consistent distributional gap: LLMs over-index on bridge-like opportunities and synthesis methods, while human researchers spread more broadly across framing and contribution types. The result suggests current LLMs produce reasonable but systematically narrower and shifted ideation relative to human researchers.