Home Knowledge Base Self-Distilled

Self-Distilled

No mentions found

This entity hasn't been tracked yet, or Iris is still building its knowledge base.

Related Articles from SNS

Physics-Guided Policy Optimization with Self-Distillation

Announce Type: new Abstract: Self-distilled policy optimization (SDPO) has become a popular paradigm for LLM post-training, where a model learns from its own predictions conditioned on privileged information. SDPO, however, is sensitive to how much each update step should be trusted: corrections from a self-teacher can be highly informative on some batches and misleading on others, and applying them uniformly with a fixed step size can destabilize training. Drawing inspiration from...

arXiv CS 7d ago

Self-Distilled Policy Gradient

arXiv:2606.04036v1 Announce Type: new Abstract: On-policy self-distillation, where a language model conditions on privileged context to supervise its own generations, is a promising source of dense supervision for sparse-reward reinforcement learning. Actually, it can be instantiated as an auxiliary full-vocabulary student-to-teacher reverse Kullback-Leibler divergence loss.

arXiv CS 6d ago

OGLS-SD: On-Policy Self-Distillation with Outcome-Guided Logit Steering for LLM Reasoning

Announce Type: replace Abstract: We study on-policy self-distillation (OPSD), where a language model improves its reasoning ability by distilling privileged teacher distributions along its own on-policy trajectories. Despite its promise, OPSD can suffer from training instability due to a pattern mismatch between teacher and student responses. Self-reflected teacher responses may introduce reflection-induced biases and response templates that miscalibrate token-level supervision, ultimately...

arXiv CS 8d ago

SHRED: Retain-Set-Free Unlearning via Self-Distillation with Logit Demotion

Announce Type: replace Abstract: Machine unlearning for large language models (LLMs) aims to selectively remove memorized content such as private data, copyrighted text, or hazardous knowledge, without costly full retraining. Most existing methods require a retain set of curated examples to prevent catastrophic degradation of general model utility, creating an extra data dependency that complicates deployment. We propose SHRED (Self-distillation via High-surprisal-only Retain-set-free...

arXiv CS 5d ago

PBSD: Privileged Bayesian Self-Distillation for Long-Horizon Credit Assignment

arXiv:2606.09348v1 Announce Type: new Abstract: Long-horizon agentic tasks pose a fundamental credit assignment challenge for outcome-base reinforcement learning: trajectory-level rewards verify final correctness but provide limited guidance on which intermediate reasoning steps or tool interactions contribute to the outcome. The difficulty is especially pronounced in multi-turn search agents, where successful trajectories may contain misleading actions and failed trajectories may contain...

arXiv CS 1d ago

MIPIC: Matryoshka Representation Learning via Self-Distilled Intra-Relational and Progressive Information Chaining

arXiv:2604.24374v2 Announce Type: replace Abstract: Representation learning is fundamental to NLP, but building embeddings that work well at different computational budgets is challenging. Matryoshka Representation Learning (MRL) offers a flexible inference paradigm through nested embeddings; however, learning such structures requires explicit coordination of how information is arranged across embedding dimensionality and model depth. In this work, we propose MIPIC (Matryoshka Representation...

arXiv CS 7d ago

Learning Visual Spatial Planning from Symbolic State via Modality-Gap-Aware Self-Distillation

Announce Type: replace Abstract: While vision-language models excel at general multimodal understanding, they still struggle with visual spatial planning. We attribute this to a perception-reasoning modality gap: visual planning requires models to infer latent state structures from pixels and then reason over the recovered structure to produce valid actions, whereas symbolic planning directly leverages explicit objects and constraints. This creates dual bottlenecks in visual state recovery...

arXiv CS 1d ago

Post-Trained MoE Can Skip Half Experts via Self-Distillation

arXiv:2605.18643v2 Announce Type: replace Abstract: Mixture-of-Experts (MoE) scales language models efficiently through sparse expert activation, and its dynamic variant further reduces computation by adjusting the activated experts in an input-dependent manner. Existing dynamic MoE methods usually rely on pre-training from scratch or task-specific adaptation, leaving the practical conversion of fully trained MoE underexplored. Enabling such adaptation would directly alleviate the inference...

arXiv CS 1d ago

Learning Visual Spatial Planning from Symbolic State via Modality-Gap-Aware Self-Distillation

Announce Type: new Abstract: While vision-language models excel at general multimodal understanding, they still struggle with visual spatial planning. We attribute this to a perception-reasoning modality gap: visual planning requires models to infer latent state structures from pixels and then reason over the recovered structure to produce valid actions, whereas symbolic planning directly leverages explicit objects and constraints. This creates dual bottlenecks in visual state recovery and...

arXiv CS 5d ago

Vision-OPD: Learning to See Fine Details for Multimodal LLMs via On-Policy Self-Distillation

arXiv:2605.18740v4 Announce Type: replace Abstract: Multimodal Large Language Models (MLLMs) still struggle with fine-grained visual understanding, where answers often depend on small but decisive evidence in the full image. We observe a regional-to-global perception gap: the same MLLM answers fine-grained questions more accurately when conditioned on evidence-centered crops than on the corresponding full images, suggesting that many failures stem from difficulty to focus on relevant...

arXiv CS 7d ago