Shourov Joarder and three co-authors propose training language models with two forms of self-generated feedback to make reinforcement learning more stable without external supervision. The approach uses reinforcement learning from internal feedback (RLIF), in which a model generates its own training signals, whereas reinforcement learning with verifiable rewards (RLVR) often uses human annotations or gold-standard solutions. The authors say RLIF methods relying on a single reward can be vulnerable to reward hacking, entropy collapse and weakened reasoning structure. The proposed method pairs an answer-level signal based on cluster voting with a completion-level signal based on token-by-token self-certainty, and GDPO-based normalization balances the scales of the two rewards. A separate technique, KL-Cov regularization, targets low-entropy token distributions and is intended to preserve exploration and prevent collapse late in training. On mathematical reasoning and code-generation benchmarks, the authors report greater stability and robustness than earlier unsupervised reinforcement-learning approaches, with results approaching those of supervised RLVR. The authors plan to release the code.