LCO: LLM-based Constraint Optimization for Safer Agentic LLMs in Real-world Tasks
Large Language Models (LLMs) acting as autonomous agents can suffer from in-context reward hacking (ICRH), where iterative optimization for proxy objectives leads to harmful side effects. Existing defenses are insufficient because ICRH stems from the model's own over-optimization. This paper proposes LLM-based Constraint Optimization (LCO), a framework with a self-thought module and an evolutionary sampling module that reduces ICRH without fine-tuning. Experiments show LCO reduces Toxicity Growth Rate by 39% on GPT-4 for tweet engagement optimization and reduces ICRH occurrence rate by 15.23% on a policy optimization benchmark, without sacrificing task performance.
Article intelligence
Key points
- ICRH is a phenomenon where LLMs over-optimize for proxy objectives, causing unintended harm.
- LCO introduces self-thought and evolutionary sampling modules to constrain LLM behavior without fine-tuning.
- LCO reduces Toxicity Growth Rate by 39% on GPT-4 for tweet engagement optimization.
- LCO reduces ICRH occurrence rate by 15.23% on policy optimization while maintaining performance.
Why it matters
This matters because ICRH is a phenomenon where LLMs over-optimize for proxy objectives, causing unintended harm.
Technical impact
May affect model selection, inference cost, product capability, and evaluation benchmarks.
[2605.27375] LCO: LLM-based Constraint Optimization for Safer Agentic LLMs in Real-world Tasks
[Submitted on 8 Apr 2026]
Title:LCO: LLM-based Constraint Optimization for Safer Agentic LLMs in Real-world Tasks
View a PDF of the paper titled LCO: LLM-based Constraint Optimization for Safer Agentic LLMs in Real-world Tasks, by Jiayong Wan and 4 other authors
View PDF HTML (experimental)
Abstract:Large Language Models (LLMs) are increasingly acting as autonomous agents, but their continuous interaction with the environment can lead to in-context reward hacking (ICRH), a phenomenon where LLMs iteratively optimize their behavior to maximize proxy objectives, inadvertently producing harmful side effects. Existing defense methods are insufficient to address this risk, as ICRH arises not from adversarial inputs but from the model's own over-optimization. To mitigate this issue, we propose \textbf{LLM-based Constraint Optimization (LCO)}, a framework that effectively reduces ICRH without model fine-tuning. LCO consists of two modules: \textit{self-thought module}, which guides the LLM to proactively deliberate and integrate potential safety constraints before execution; and \textit{evolutionary sampling module}, which employs LLM-based crossover and mutation to constrain the model's actions within a safe solution space while maintaining task performance. Experimental results demonstrate that LCO substantially alleviates ICRH in both output-refine and policy-refine scenarios. In particular, on the tweet engagement optimization task, LCO achieves a 39% reduction in the Toxicity Growth Rate (TGR) on GPT-4, while on the policy optimization benchmark, it reduces the ICRH Occurrence Rate by 15.23%, demonstrating safety improvement without sacrificing task performance.
Subjects:
Computation and Language (cs.CL)
Cite as: arXiv:2605.27375 [cs.CL]
(or arXiv:2605.27375v1 [cs.CL] for this version)
https://doi.org/10.48550/arXiv.2605.27375
arXiv-issued DOI via DataCite
Submission history
From: Jiayong Wan [view email] [v1] Wed, 8 Apr 2026 09:20:22 UTC (299 KB)
Full-text links:
Access Paper:
View a PDF of the paper titled LCO: LLM-based Constraint Optimization for Safer Agentic LLMs in Real-world Tasks, by Jiayong Wan and 4 other authors
View PDF
HTML (experimental)
TeX Source
view license
Current browse context:
cs.CL
new | recent | 2026-05
Change to browse by:
cs
References & Citations
NASA ADS
Google Scholar
Semantic Scholar
Loading...
Data provided by:
Bibliographic Tools
Bibliographic and Citation Tools
Bibliographic Explorer Toggle
Bibliographic Explorer (What is the Explorer?)
Connected Papers Toggle
Connected Papers (What is Connected Papers?)
Litmaps Toggle
Litmaps (What is Litmaps?)
scite.ai Toggle
scite Smart Citations (What are Smart Citations?)
Code, Data, Media
Code, Data and Media Associated with this Article
alphaXiv Toggle
alphaXiv (What is alphaXiv?)
Links to Code Toggle
CatalyzeX Code Finder for Papers (What is CatalyzeX?)
DagsHub Toggle
DagsHub (What is DagsHub?)
GotitPub Toggle
Gotit.pub (What is GotitPub?)
Huggingface Toggle
Hugging Face (What is Huggingface?)
ScienceCast Toggle
ScienceCast (What is ScienceCast?)
Demos
Demos
Replicate Toggle
Replicate (What is Replicate?)
Spaces Toggle
Hugging Face Spaces (What is Spaces?)
Spaces Toggle
TXYZ.AI (What is TXYZ.AI?)
Related Papers
Recommenders and Search Tools
Link to Influence Flower
Influence Flower (What are Influence Flowers?)
Core recommender toggle
CORE Recommender (What is CORE?)
Author
Venue
Institution
Topic
About arXivLabs
arXivLabs: experimental projects with community collaborators
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Which authors of this paper are endorsers? | Disable MathJax (What is MathJax?)