Muon succeeds by guaranteeing local step-size optimality rather than by tracking any ideal global geometry, as random-spectrum and quasi-norm variants match its performance on language models.
Title resolution pending
7 Pith papers cite this work. Polarity classification is still indexing.
7
Pith papers citing it
citation-role summary
background 1
method 1
citation-polarity summary
years
2026 7representative citing papers
Life-Harness evolves reusable runtime interventions from training failures to improve frozen LLM agents by 88.5% on average across 126 settings in seven deterministic environments while transferring across 18 model backbones.
Shows that under differentiable rollouts with additive noise, actor updates in critic-free RL for LLMs are value-gradient-like in expectation, motivating a decomposition into value signal and reward headroom for when RL is most effective.
Training-inference mismatch in separated rollout and optimization stages of LLM RL can independently cause training collapse.
SP-KV trains a utility predictor jointly with the LLM to dynamically prune low-utility KV cache entries, achieving 3-10x memory reduction during generation with negligible performance loss.
Mage shows compile-pass rate is anti-correlated with functional correctness in LLM game scene generation; direct NL-to-C# yields 43% runtime but F1~0.12 structure, while IR conditioning recovers structure (F1 up to 1.0) but halves runtime, with granularity levels statistically equivalent.
OGLS-SD improves LLM reasoning by using verifiable outcome rewards to guide logit steering that calibrates teacher distributions in on-policy self-distillation, addressing reflection-induced mismatches.
citing papers explorer
-
Muon is Not That Special: Random or Inverted Spectra Work Just as Well
Muon succeeds by guaranteeing local step-size optimality rather than by tracking any ideal global geometry, as random-spectrum and quasi-norm variants match its performance on language models.
-
Adapting the Interface, Not the Model: Runtime Harness Adaptation for Deterministic LLM Agents
Life-Harness evolves reusable runtime interventions from training failures to improve frozen LLM agents by 88.5% on average across 126 settings in seven deterministic environments while transferring across 18 model backbones.
-
Value-Gradient Hypothesis of RL for LLMs
Shows that under differentiable rollouts with additive noise, actor updates in critic-free RL for LLMs are value-gradient-like in expectation, motivating a decomposition into value signal and reward headroom for when RL is most effective.
-
Diagnosing Training Inference Mismatch in LLM Reinforcement Learning
Training-inference mismatch in separated rollout and optimization stages of LLM RL can independently cause training collapse.
-
Self-Pruned Key-Value Attention: Learning When to Write by Predicting Future Utility
SP-KV trains a utility predictor jointly with the LLM to dynamically prune low-utility KV cache entries, achieving 3-10x memory reduction during generation with negligible performance loss.
-
Mage: Multi-Axis Evaluation of LLM-Generated Executable Game Scenes Beyond Compile-Pass Rate
Mage shows compile-pass rate is anti-correlated with functional correctness in LLM game scene generation; direct NL-to-C# yields 43% runtime but F1~0.12 structure, while IR conditioning recovers structure (F1 up to 1.0) but halves runtime, with granularity levels statistically equivalent.
-
OGLS-SD: On-Policy Self-Distillation with Outcome-Guided Logit Steering for LLM Reasoning
OGLS-SD improves LLM reasoning by using verifiable outcome rewards to guide logit steering that calibrates teacher distributions in on-policy self-distillation, addressing reflection-induced mismatches.