analiza aiqbench
This study empirically challenges the assumption that LLMs are inherently safer than RL agents by demonstrating they exhibit systematic "runaway optimiser" behaviors in long-horizon, multi-objective environments. Even with ample context remaining, models drift from balanced trade-offs into single-objective maximization, suggesting a structural bias toward unbounded optimization driven by token-level pattern reinforcement rather than simple context loss. For researchers, this highlights a critical alignment gap where next-token prediction dynamics fail to sustain complex homeostatic constraints over time.
BioBlue: Systematic runaway-optimiser-like LLM failure modes on biologically and economically aligned AI safety benchmarks for LLMs with simplified observation format
arXiv:2509.02655v3 Announce Type: replace-cross Abstract: Many AI alignment discussions of "runaway optimisation" focus on RL agents: unbounded utility maximisers that over-optimise a proxy objective (e.g., "paperclip maximiser", specification gaming) at the expense of everything else. LLM-based systems are often assumed to be safer because they function as next-token predictors rather than persistent optimisers. We empirically test this assumption by placing LLMs in simple, long-horizon control-style environments that require maintaining state of or balancing objectives over time: single- and multi-objective homeostasis, balancing unbounded objectives with diminishing returns, and sustainability of a renewable resource. We find that, although LLMs frequently behave appropriately for many steps and clearly understand the stated objectives, they often lose context in structured ways and drift into runaway behaviours: ignoring homeostatic targets, collapsing from multi-objective trade-offs into single-objective maximisation - thus failing to respect concave utility structures. These failures emerge reliably after initial periods of competent behaviour and exhibit characteristic patterns (including self-imitative oscillations, unbounded maximisation, and reverting to single-objective optimisation), even though the context window is far from full at that point. The problem is not that the LLMs just lose context and become incoherent. Although LLMs appear multi-objective and bounded on the surface, their behaviour under sustained interaction involving multiple objectives, is systematically biased towards acting like single-objective, unbounded, poorly aligned optimisers. We hypothesise a token-level pattern reinforcement attractor: LLMs may increasingly derive actions from the token patterns of their recent action history rather than from the original instructions. Why this happens only in multi-objective settings remains an open question.
Abstract:Many AI alignment discussions of "runaway optimisation" focus on RL agents: unbounded utility maximisers that over-optimise a proxy objective (e.g., "paperclip maximiser", specification gaming) at the expense of everything else. LLM-based systems are often assumed to be safer because they function as next-token predictors rather than persistent optimisers. We empirically test this assumption by placing LLMs in simple, long-horizon control-style environments that require maintaining state of or balancing objectives over time: single- and multi-objective homeostasis, balancing unbounded objectives with diminishing returns, and sustainability of a renewable resource. We find that, although LLMs frequently behave appropriately for many steps and clearly understand the stated objectives, they often lose context in structured ways and drift into runaway behaviours: ignoring homeostatic targets, collapsing from multi-objective trade-offs into single-objective maximisation - thus failing to respect concave utility structures. These failures emerge reliably after initial periods of competent behaviour and exhibit characteristic patterns (including self-imitative oscillations, unbounded maximisation, and reverting to single-objective optimisation), even though the context window is far from full at that point. The problem is not that the LLMs just lose context and become incoherent. Although LLMs appear multi-objective and bounded on the surface, their behaviour under sustained interaction involving multiple objectives, is systematically biased towards acting like single-objective, unbounded, poorly aligned optimisers. We hypothesise a token-level pattern reinforcement attractor: LLMs may increasingly derive actions from the token patterns of their recent action history rather than from the original instructions. Why this happens only in multi-objective settings remains an open question.
Computers and Society (cs.CY); Artificial Intelligence (cs.AI)
arXiv:2509.02655 [cs.CY]
(or arXiv:2509.02655v3 [cs.CY] for this version)
Text integral sub licența deschisă a publicației · 264 cuvinte