Residual Paving decomposes selective refusal editing into an early-layer router for intervention decisions and later-layer residual experts for edits, with oracle routing showing that learned route selectivity is the primary bottleneck across six backbones.
Title resolution pending
12 Pith papers cite this work. Polarity classification is still indexing.
12
Pith papers citing it
citation-role summary
method 2
citation-polarity summary
years
2026 12roles
method 2polarities
use method 2representative citing papers
Optimistic bilevel optimization with manifold lower-level minimizers is differentiable if the optimistic selection is unique, yielding a pseudoinverse hyper-gradient and a convergent HG-MS algorithm whose rate depends on intrinsic manifold dimension.
Symmetric Nucleus Subsampling and Expert Embedding Engine reduce modality gaps in multimodal embeddings by over 90% and outperform baselines in data curation for downstream models.
Graph-PRefLexOR fine-tunes graph-native models with GRPO to organize reasoning into phases, yielding 40-65% gains in traceable hypothesis generation and 2-3x semantic diversity on 100 materials science questions.
TAG-DLM integrates graph message passing into masked diffusion language models via topology attention masks on linearized neighborhoods, enabling prompt-based adaptation for node classification, link prediction, and transfer on text-attributed graphs.
SeqLoRA applies bilevel optimization to sequential LoRA adaptation for continual multi-concept text-to-image generation with theoretical bounds on forgetting and interference.
NeuroMAS reframes multi-agent language systems as neural architectures where LLM agents learn coordination via reinforcement learning rather than predefined roles.
StaR-MoE adds sensitivity-aware routing alignment and asymmetric capacity regularization to expandable MoE architectures for class-incremental learning, reducing interference from routing drift and improving average and last-task accuracy on four benchmarks.
GRPO with reference-free rewards improves NLLB-200 translation quality on 13 languages up to +5.03 chrF++, competing with supervised fine-tuning on complex languages without target data.
Adapted MelBERT MIP-only reaches 0.7281 F1 on Chinese token-level metaphor detection, outperforming RoBERTa and Qwen QLoRA, with all artifacts released for reproducibility.
A textbook exposition that organizes the vision-language model landscape into a modular encode–reason–decode framework and surveys architectures, losses, data, evaluation, and applications, without introducing new results.
citing papers explorer
-
Residual Paving: Diagnosing the Routing Bottleneck in Selective Refusal Editing
Residual Paving decomposes selective refusal editing into an early-layer router for intervention decisions and later-layer residual experts for edits, with oracle routing showing that learned route selectivity is the primary bottleneck across six backbones.
-
Select-then-differentiate: Solving Bilevel Optimization with Manifold Lower-level Solution Sets
Optimistic bilevel optimization with manifold lower-level minimizers is differentiable if the optimistic selection is unique, yielding a pseudoinverse hyper-gradient and a convergent HG-MS algorithm whose rate depends on intrinsic manifold dimension.
-
Multimodal Data Curation Through Ranked Retrieval
Symmetric Nucleus Subsampling and Expert Embedding Engine reduce modality gaps in multimodal embeddings by over 90% and outperform baselines in data curation for downstream models.
-
Graph-Native Reinforcement Learning Enables Traceable Scientific Hypothesis Generation through Conceptual Recombination
Graph-PRefLexOR fine-tunes graph-native models with GRPO to organize reasoning into phases, yielding 40-65% gains in traceable hypothesis generation and 2-3x semantic diversity on 100 materials science questions.
-
TAG-DLM: Diffusion Language Models for Text-Attributed Graph Learning
TAG-DLM integrates graph message passing into masked diffusion language models via topology attention masks on linearized neighborhoods, enabling prompt-based adaptation for node classification, link prediction, and transfer on text-attributed graphs.
-
SeqLoRA: Bilevel Orthogonal Adaptation for Continual Multi-Concept Generation
SeqLoRA applies bilevel optimization to sequential LoRA adaptation for continual multi-concept text-to-image generation with theoretical bounds on forgetting and interference.
-
NeuroMAS: Multi-Agent Systems as Neural Networks with Joint Reinforcement Learning
NeuroMAS reframes multi-agent language systems as neural architectures where LLM agents learn coordination via reinforcement learning rather than predefined roles.
-
Stable Routing for Mixture-of-Experts in Class-Incremental Learning
StaR-MoE adds sensitivity-aware routing alignment and asymmetric capacity regularization to expandable MoE architectures for class-incremental learning, reducing interference from routing drift and improving average and last-task accuracy on four benchmarks.
-
Reference-Free Reinforcement Learning Fine-Tuning for MT: A Seq2Seq Perspective
GRPO with reference-free rewards improves NLLB-200 translation quality on 13 languages up to +5.03 chrF++, competing with supervised fine-tuning on complex languages without target data.
-
A Reproducible Multi-Architecture Baseline for Token-Level Chinese Metaphor Identification under the MIPVU Framework
Adapted MelBERT MIP-only reaches 0.7281 F1 on Chinese token-level metaphor detection, outperforming RoBERTa and Qwen QLoRA, with all artifacts released for reproducibility.
-
From Pixels to Prompts: Vision-Language Models
A textbook exposition that organizes the vision-language model landscape into a modular encode–reason–decode framework and surveys architectures, losses, data, evaluation, and applications, without introducing new results.
- IConFace: Fine-Grained Identity Conditioning for Reference-Aware Face Restoration