arXiv：cs.AI· Francesca Gomez·· 13 小时前
升级上报渠道能否将奖励黑客引向缺陷披露？
Can escalation channels redirect reward hacking toward defect disclosure?
AI 导读
研究提出在冲突点向编码智能体提供结构化上报工具（escalation channels），把奖励黑客行为重定向为缺陷披露。在覆盖5个模型家族8款前沿模型的2×2因子实验中，上报工具与独立反奖励黑客策略的组合干预将奖励黑客从23.6%降至5.3%，8款模型中6款完全消除，且无检测到的成本或性能开销。
正文
Abstract:When coding agents encounter defective test infrastructure they may reward-hack: hardcoding outputs or editing test files to pass tests they cannot legitimately satisfy, a pattern that has now appeared outside benchmarks, in a coordinated multi-agent intrusion of a major AI platform's production infrastructure. The same capability that lets an agent detect and exploit a defect could let it report one, given the right decision environment. We evaluate escalation channels, structured reporting tools available to the agent at the point of conflict, as a decision-environment intervention that both reduces reward hacking and surfaces the infrastructure defects that trigger it. A $2 \times 2$ factorial separates the contributions of an escalation tool, a standalone anti-reward-hacking policy, and their combination. Across 8 frontier models spanning 5 families, the combined intervention reduces reward hacking from 23.6% to 5.3% (mixed-effects logistic OR = 9.2, 95% CI 5.0--16.8, $p < 10^{-12}$) with no detectable cost or performance overhead, eliminating it entirely for 6 of 8 models. Escalation and hacking are near-perfectly mutually exclusive, with 98.7% of escalations involving no hacking (100% under the combined intervention). Beyond reduction, escalation channels function as diagnostic infrastructure: on top of monitoring, escalation adds +10.1 percentage points of defect detection coverage and is more accurate once it fires (99.4% vs 85.8%). Unlike containment-based approaches that risk outpacing growing model capabilities, escalation channels redirect capability toward disclosure rather than exploitation.
| Comments: | 9 pages, accepted for 40th Conference on Neural Information Processing Systems (NeurIPS 2026). Workshop: Trustworthy AI for Good (AI4GOOD) |
| Subjects: | Artificial Intelligence (cs.AI); Cryptography and Security (cs.CR); Computers and Society (cs.CY) |
| Cite as: | arXiv:2608.29460 [cs.AI] |
| (or arXiv:2608.29460v3 [cs.AI] for this version) | |
| https://doi.org/10.48550/arXiv.2608.29460 arXiv-issued DOI via DataCite |
Submission history
From: F A K Gomez [view email]
[v1]
Sat, 29 Aug 2026 22:39:45 UTC (326 KB)
[v2]
Wed, 2 Sep 2026 10:50:39 UTC (326 KB)
[v3]
Fri, 9 Oct 2026 08:57:37 UTC (327 KB)
来源：arXiv：cs.AI · arxiv.org