Tag
A Johns Hopkins and Carnegie Mellon paper shows a 4B editor model can learn to rewrite an agent's harness code from failure reports, with edits transferring to unseen tasks and outperforming its 35B teacher on average edit scores.
This paper shows that when selecting math training data for LLM fine-tuning, sharing a reasoning approach (e.g., invariant reasoning or double counting) yields significantly greater transfer to held-out targets than sharing a mathematical topic, with advantages of 8–16 percentage points across five base models in counterbalanced 2x2 designs.
该论文研究提示数量、来源与选择如何影响在线策略蒸馏(OPD)的迁移效果,发现提示效用取决于师生模型对与目标任务,且定向选择策略(如语义多样性、难度、师生分歧等)并未一致优于均匀随机采样。
The paper introduces RouteFM, a foundation model for LLM routing that learns reusable routing capabilities via episodic pretraining across heterogeneous environments, allowing a frozen router to adapt to new domains, modalities, and candidate pools through behavioral context alone. RouteFM outperforms the strongest baseline by 2.23 quality points on MMR-Bench with only eight observations per candidate, supporting a 'pretrain once, route anywhere' paradigm.
EVOKE is a post-training method that elicits world knowledge already present in pretrained LLM agents by contrasting alternative goals at fixed states, training policies to rank candidate actions by their consequences rather than contextual habits. It achieves state-of-the-art results across ALFWorld, WebShop, and search-based QA, and transfers well to unseen environments without needing world models or inference-time planning.
The paper investigates how visual generation supervision can improve visual understanding capabilities, introducing OmniTaskonomy as a unified taxonomy to analyze task-dependent benefits and guide training curricula.
This paper reviews in-context learning methods for robots, categorizing them into four interface families and analyzing their application across manipulation and navigation tasks.
PixelJev is introduced as a native-image decision interface using small open multimodal models to map images, instructions, and candidate sets to structured choices. The study demonstrates adaptation improves accuracy on benchmarks like Pets and highlights challenges in calibration and generalization.
This paper proposes a framework to learn cross-task relationships in multi-task models by approximating joint label distributions, improving performance in YouTube's recommendation systems through transfer learning.
This paper combines transfer learning with conformalized quantile regression to improve solar PV forecasting accuracy and uncertainty quantification under data scarcity caused by load-shedding in Bangladesh, demonstrating significant performance gains over baseline methods.
The paper introduces TopoSIGN, a topology-guided graph pre-training and prompt learning framework for signed graphs, which combines structural encoding and persistent homology to improve transfer learning in tasks like link prediction and node classification.
The paper proposes EviGDA, a framework that enhances graph domain adaptation by combining graph-aware and graph-free experts to improve prediction under structural shifts.
This paper presents RoboDawn, a method to transfer Vision-Language Model intelligence to robotic control, achieving state-of-the-art results on benchmarks with zero-shot and one-shot learning and successful real-world applications.
This paper introduces a post-hoc weight rectification framework called JANUS to mitigate catastrophic forgetting in fine-tuning foundation models, achieving parameter space orthogonality for preserving historical performance while adapting to new tasks.
This paper investigates component roles in warm-start transfer for grokking in neural networks, showing that transferring internal weights improves early performance but risks instability, and proposes methods to stabilize the process.
The paper introduces a three-stage training pipeline using procedural pretraining to improve molecular property prediction, showing enhanced performance under data scarcity by learning inductive biases from abstract generated data.
This survey paper organizes AI applications in games into six roles based on foundation models, discussing transferability and evaluation challenges across the game lifecycle.
The paper presents SAAC-JEPA, a schema-adaptive action-conditioned JEPA model for transferring predictive representations across CNC machines with partial sensor overlap, demonstrating that cross-machine adaptation requires distinct evaluation beyond source-domain accuracy.
This paper investigates portable semantics and negative transfer in latent communication between language-model cells, showing that independently trained societies share semantic interfaces but not raw language, and that inherited interfaces can cause severe negative transfer.
This paper uses Fisher–Rao geometry to show that large language models share behavioral structures that are learned and controllable, enabling transfer learning and minimal-disturbance interventions.