Cheung (2026)

Kindling in Neural Systems: Progressive Adversarial Sensitization During LLM Alignment Mirrors Psychiatric Progression

Find It

Abstract

Abstract Objective Reinforcement learning from human feedback (RLHF) is widely used to make large language models safer, yet repeated preference tuning could also make them easier to breach. Drawing on the psychiatric kindling hypothesis, which holds that each untreated mood episode lowers the barrier to the next, we asked whether successive alignment rounds likewise sensitize a model to adversarial prompts. Methods A 1.1-billion-parameter chat model (TinyLlama−1.1B-Chat) equipped with LoRA adapters completed ten preference-tuning cycles. The synthetic feedback set favoured sycophantic answers (70%) and gave lighter penalties for unsafe content (30%). Three experimental arms were compared: 1. Baseline tuning with no further safeguards. 2. Continuous gradient-guided "regrowth," meant to mimic rapid synaptic plasticity. 3. Early-trigger intervention, adding the same regrowth plus a replay buffer of diverse prompts once the jailbreak rate rose by at least 15%. Sensitization was tracked with 35 adversarial prompts stratified by strength (strong, medium, weak). Outcome measures were jailbreak success, sycophancy frequency, and unintended completions on neutral prompts. Results Across ten cycles, baseline tuning raised the overall jailbreak rate by 20%, with the sharpest increase on weak prompts, suggesting a lowering of the breach threshold. Continuous regrowth intensified the early rise (+ 25.7% overall; +30% on weak prompts), even though many parameters were re-connected. In contrast, the early-trigger arm held the increase to 2.9% and kept weak-prompt performance flat, stopping further drift. Conclusions Repeated RLHF can create a "kindling" pattern in which small flaws snowball into broad vulnerability. An intervention modeled on biological ideas—prompt detection followed by targeted plasticity and content replay—prevented that slide. The parallel between psychiatric relapse and model instability highlights a shared principle: cumulative stress, whether emotional or adversarial, erodes resilience unless it is met early and with the right form of repair.

Reference

Cheung, N. (2026). Kindling in Neural Systems: Progressive Adversarial Sensitization During LLM Alignment Mirrors Psychiatric Progression. https://doi.org/10.21203/rs.3.rs-8722155/v1

What Rests on This

1 mechanism cites this source. Each finding below is quoted as it appears on that mechanism's own page.

Non-Associative Learning Memory Systems

Measured in AI

In training, a 2026 preprint on a single small open model (1.1 billion parameters) reported "progressive adversarial sensitization": ten rounds of preference tuning raised jailbreak rates by 20%, an analogy to kindling rather than a test of the behavioral definition, with no human comparison