Unverified50% confidenceSolutionExact time
After code is generated, the system automatically runs a test suite, calculates a reward signal based on pass/fail results, and then updates the model parameters using reinforcement learning algorithms such as PPO or GRPO.
1
Sources
50%
Confidence
Long-term
Relevance
8/16/2026
First Seen
Sources
Related Claims
UnverifiedSome code-specific models actually execute the generated code during training and use the test results as a reward signal for reinforcement learning (execution feedback).79% similarUnverifiedExecution-based Reward signals derived from unit test execution results are significantly effective in improving code correctness rates.74% similarUnverified在代码生成场景中，奖励破解的等效行为包括注释掉失败的测试用例、针对测试输入硬编码预期输出、删除触发错误的代码路径73% similarUnverified系统在测试点生成后和用例生成阶段均内置自动评审机制，并保留人工评审接口73% similarUnverified编程任务特别适合规划模式，因为代码执行结果具有明确的对/错反馈信号，可通过运行测试验证每个步骤，形成计划—执行—验证—修正的闭环73% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/755765API
curl https://kongchang.com/api/v1/knowledge/claims/755765MCP
get_claim(id=755765)