AGENTIC-SYSTEMSApril 3, 20261 min read30 views

How Do Language Models Process Ethical Instructions? Deliberation, Consistency, and Other-Recognition Across Four Models

A study of four major LLMs reveals that internal ethical processing varies significantly across models, suggesting that surface-level compliance with safety instructions does not always correlate with genuine ethical reasoning.

Computer Science > Computation and Language

Title:How Do Language Models Process Ethical Instructions? Deliberation, Consistency, and Other-Recognition Across Four Models

View PDF HTML (experimental)Abstract:Alignment safety research assumes that ethical instructions improve model behavior, but how language models internally process such instructions remains unknown. We conducted over 600 multi-agent simulations across four models (Llama 3.3 70B, GPT-4o mini, Qwen3-Next-80B-A3B, Sonnet 4.5), four ethical instruction formats (none, minimal norm, reasoned norm, virtue framing), and two languages (Japanese, English). Confirmatory analysis fully replicated the Llama Japanese dissociation pattern from a prior study ($\mathrm{BF}_{10} > 10$ for all three hypotheses), but none of the other three models reproduced this pattern, establishing it as model-specific. Three new metrics -- Deliberation Depth (DD), Value Consistency Across Dilemmas (VCAD), and Other-Recognition Index (ORI) -- revealed four distinct ethical processing types: Output Filter (GPT; safe outputs, no processing), Defensive Repetition (Llama; high consistency through formulaic repetition), Critical Internalization (Qwen; deep deliberation, incomplete integration), and Principled Consistency (Sonnet; deliberation, consistency, and other-recognition co-occurring). The central finding is an interaction between processing capacity and instruction format: in low-DD models, instruction format has no effect on internal processing; in high-DD models, reasoned norms and virtue framing produce opposite effects. Lexical compliance with ethical instructions did not correlate with any processing metric at the cell level ($r = -0.161$ to $+0.256$, all $p > .22$; $N = 24$; power limited), suggesting that safety, compliance, and ethical processing are largely dissociable. These processing types show structural correspondence to patterns observed in clinical offender treatment, where formal compliance without internal processing is a recognized risk signal.

Source: arXiv cs.AI Recent

More in this category

agentic-systems

VeriSimpl: Robust Optimization Modeling from Natural Language using Simplification-based Verification

Researchers have introduced VeriSimpl, a novel framework that leverages LLMs and simplification-based verification to accurately translate natural language descriptions into executable optimization formulations.

agentic-systems

Semi-Supervised Text-Attributed Graph Distillation

Researchers propose a unified semi-supervised framework guided by Wasserstein Distance to tackle scalability bottlenecks in Text-Attributed Graphs (TAGs). The approach achieves a state-of-the-art trade-off between performance and data compression for both GNN- and LLM-based downstream tasks.

agentic-systems

Benchmarking the Personalization Capabilities of Large Language Models

A new study introduces SDR-Bench and SDR-Arena to benchmark the personalization capabilities of LLMs in two-party persuasion scenarios, revealing a personalization plateau among frontier models.

agentic-systems

SonicSampler: Unified Tile-Aware Kernels for LLM Sampling and Speculative Verification

Researchers introduced SonicSampler, a unified suite of tile-aware Triton kernels designed to accelerate LLM sampling and speculative verification. By fusing the entire pipeline into a single CUDA Graph-compatible kernel, SonicSampler achieves up to 16x speedups over state-of-the-art baselines.

agentic-systems

Enabling Scalable Topology Inference in Distribution Systems via Constrained Multi-Source Inference

A new research study introduces a constrained multi-source inference framework to accurately reconstruct power distribution system topology. Validated on over 8,000 smart meters in the U.S., the method achieves over 95% accuracy while drastically reducing computational overhead.

agentic-systems

Incomplete Prompt Jailbreaks in Large Language Models

Researchers have conceptualized Incomplete Prompt Jailbreaks (IPJ), a vulnerability where incomplete harmful prompts bypass LLM safeguards due to delayed refusal behaviors. The discovery of functional neurons offers a novel path for precise neuron-level defense interventions.