ControBench is a new interaction-aware benchmark combining heterogeneous graphs and rich text for controversial discourse analysis on social networks.
RL Fine-Tuning Heals OOD Forgetting in SFT
7 Pith papers cite this work. Polarity classification is still indexing.
7
Pith papers citing it
abstract
Supervised Fine-Tuning (SFT) followed by Reinforcement Learning (RL) is a standard post-training recipe for improving Large Language Models (LLM) reasoning, but why it works remains unclear. We revisit the common claim that ``SFT memorizes, RL generalizes'' through checkpoint-wise analyses of in-distribution (ID) and out-of-distribution (OOD) reasoning. We find that OOD performance often peaks early during SFT and then declines despite continued improvement in ID reasoning. RL typically does not surpass this early SFT peak; rather, it restores OOD capability lost during later SFT, and only from a bounded range of SFT checkpoints. Further spectral analysis shows that this forgetting-and-recovery pattern correlates with rotations of singular vectors, while singular values remain largely stable. These findings suggest a more precise view of post-training dynamics: SFT can forget, RL can recover, and controlling singular-vector rotation may improve OOD robustness. Code is available at \href{https://github.com/jinhangzhan/RL\_Heals\_SFT.git}{https://github.com/jinhangzhan/RL\_Heals\_SFT}.
citation-role summary
background 1
citation-polarity summary
roles
background 1polarities
background 1representative citing papers
CoT training in LLM agents improves prompt-action quality more than the advantage of generated reasoning, and selectively masking action supervision improves out-of-domain generalization.
Grounding-certainty-guided Adaptive Advantage and Dynamic Clipping improve continual GUI grounding over RFT baselines on ScreenSpot-V1/V2/Pro.
RPSFT improves the in-domain versus out-of-domain performance trade-off during LLM supervised fine-tuning by penalizing rotations in pretrained singular subspaces as a proxy for loss-sensitive directions.
RLVR drives a concept network in LLMs through nucleation and freezing into inverse trees that support slow thinking, and intervening with brief SFT at peak frustration outperforms standard RLVR while post-freeze SFT causes forgetting.
Excessive SFT reduces LLM plasticity for RL; Rejuvenation restores it via base-anchored fusion and targeted neuron resets, yielding better RL performance and OOD generalization.
Outcome-level RL with binary or composite rewards improves compositional generalization over supervised fine-tuning by avoiding overfitting to frequent training patterns.
citing papers explorer
-
ControBench: An Interaction-Aware Benchmark for Controversial Discourse Analysis on Social Networks
ControBench is a new interaction-aware benchmark combining heterogeneous graphs and rich text for controversial discourse analysis on social networks.
-
Where Do CoT Training Gains Land in LLM based Agents?
CoT training in LLM agents improves prompt-action quality more than the advantage of generated reasoning, and selectively masking action supervision improves out-of-domain generalization.
-
GUI-AC: Enhancing Continual Learning in GUI Agents
Grounding-certainty-guided Adaptive Advantage and Dynamic Clipping improve continual GUI grounding over RFT baselines on ScreenSpot-V1/V2/Pro.
-
Rotation-Preserving Supervised Fine-Tuning
RPSFT improves the in-domain versus out-of-domain performance trade-off during LLM supervised fine-tuning by penalizing rotations in pretrained singular subspaces as a proxy for loss-sensitive directions.
-
Emergent Slow Thinking in LLMs as Inverse Tree Freezing
RLVR drives a concept network in LLMs through nucleation and freezing into inverse trees that support slow thinking, and intervening with brief SFT at peak frustration outperforms standard RLVR while post-freeze SFT causes forgetting.
-
When RL Fails after SFT: Rejuvenating Model Plasticity for Robust SFT-to-RL Handoff
Excessive SFT reduces LLM plasticity for RL; Rejuvenation restores it via base-anchored fusion and targeted neuron resets, yielding better RL performance and OOD generalization.
-
Reinforcement Learning for Compositional Generalization with Outcome-Level Optimization
Outcome-level RL with binary or composite rewards improves compositional generalization over supervised fine-tuning by avoiding overfitting to frequent training patterns.