GRACE: Generative Representation Learning via Contrastive Policy Optimization

• URL: http://arxiv.org/abs/2510.04506v1
• Date: Mon, 06 Oct 2025 05:46:56 GMT
• Title: GRACE: Generative Representation Learning via Contrastive Policy Optimization
• Authors: Jiashuo Sun, Shixuan Liu, Zhaochen Su, Xianrui Zhong, Pengcheng Jiang, Bowen Jin, Peiran Li, Weijia Shi, Jiawei Han,
• Abstract summary: We introduce GRACE, a framework that reimagines contrastive signals as rewards for a generative policy.<n>GRACE produces explicit, human-interpretable rationales--structured natural language explanations.<n>On MTEB benchmark, GRACE yields broad cross category gains.
• Score: 44.40343928975602
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Prevailing methods for training Large Language Models (LLMs) as text encoders rely on contrastive losses that treat the model as a black box function, discarding its generative and reasoning capabilities in favor of static embeddings. We introduce GRACE (Generative Representation Learning via Contrastive Policy Optimization), a novel framework that reimagines contrastive signals not as losses to be minimized, but as rewards that guide a generative policy. In GRACE, the LLM acts as a policy that produces explicit, human-interpretable rationales--structured natural language explanations of its semantic understanding. These rationales are then encoded into high-quality embeddings via mean pooling. Using policy gradient optimization, we train the model with a multi-component reward function that maximizes similarity between query positive pairs and minimizes similarity with negatives. This transforms the LLM from an opaque encoder into an interpretable agent whose reasoning process is transparent and inspectable. On MTEB benchmark, GRACE yields broad cross category gains: averaged over four backbones, the supervised setting improves overall score by 11.5% over base models, and the unsupervised variant adds 6.9%, while preserving general capabilities. This work treats contrastive objectives as rewards over rationales, unifying representation learning with generation to produce stronger embeddings and transparent rationales. The model, data and code are available at https://github.com/GasolSun36/GRACE.

Related papers

• $\ abla$-Reasoner: LLM Reasoning via Test-Time Gradient Descent in Latent Space [71.23672814629448]
  $nabla$-Reasoner is an iterative generation framework that integrates differentiable optimization over token logits into the decoding loop.<n>$nabla$-Reasoner achieves over 20% accuracy improvement on a challenging mathematical reasoning benchmark.
  arXiv  Detail & Related papers  (2026-03-05T08:42:54Z)
• FAIL: Flow Matching Adversarial Imitation Learning for Image Generation [52.643484089126844]
  Post-training of flow matching models-aligning the output distribution with a high-quality target-is mathematically equivalent to Imitation learning.<n>We propose Flow Matching Adrial Learning (FAIL), which minimizes policy-expert divergence through adversarial training without explicit rewards or pairwise comparisons.
  arXiv  Detail & Related papers  (2026-02-12T16:36:33Z)
• Policy-Conditioned Policies for Multi-Agent Task Solving [53.67744322553693]
  In this work, we propose a paradigm shift that bridges the gap by representing policies as human-interpretable source code.<n>We reformulate the learning problem by utilizing Large Language Models (LLMs) as approximate interpreters.<n>We formalize this process as textitProgrammatic Iterated Best Response (PIBR), an algorithm where the policy code is optimized by textual gradients.
  arXiv  Detail & Related papers  (2025-12-24T07:42:10Z)
• Compile Scene Graphs with Reinforcement Learning [69.36723767339001]
  Next-token prediction is the fundamental principle for training large language models (LLMs)<n>We introduce R1-SGG, a multimodal LLM (M-LLM) initially trained via supervised fine-tuning (SFT) on the scene graph dataset.<n>We design a set of graph-centric rewards, including three recall-based variants -- Hard Recall, Hard Recall+Relax, and Soft Recall.
  arXiv  Detail & Related papers  (2025-04-18T10:46:22Z)
• Contrastive Policy Gradient: Aligning LLMs on sequence-level scores in a supervised-friendly fashion [43.77763433288893]
  We introduce Contrastive Policy Gradient, or CoPG, a simple and mathematically principled new RL algorithm that can estimate the optimal policy even from off-policy data.<n>We show this approach to generalize the direct alignment method IPO (identity preference optimization) and classic policy gradient.<n>We experiment with the proposed CoPG on a toy bandit problem to illustrate its properties, as well as for finetuning LLMs on a summarization task.
  arXiv  Detail & Related papers  (2024-06-27T14:03:49Z)
• Fine-Tuning Language Models with Reward Learning on Policy [68.70065254564642]
  Reinforcement learning from human feedback (RLHF) has emerged as an effective approach to aligning large language models (LLMs) to human preferences. Despite its popularity, (fixed) reward models may suffer from inaccurate off-distribution. We propose reward learning on policy (RLP), an unsupervised framework that refines a reward model using policy samples to keep it on-distribution.
  arXiv  Detail & Related papers  (2024-03-28T10:02:10Z)
• Stackelberg Batch Policy Learning [3.5426153040167754]
  Batch reinforcement learning (RL) defines the task of learning from a fixed batch of data lacking exhaustive exploration. Worst-case optimality algorithms, which calibrate a value-function model class from logged experience, have emerged as a promising paradigm for batch RL. We propose a novel gradient-based learning algorithm: StackelbergLearner, in which the leader player updates according to the total derivative of its objective instead of the usual individual gradient.
  arXiv  Detail & Related papers  (2023-09-28T06:18:34Z)
• Positive-Negative Equal Contrastive Loss for Semantic Segmentation [8.664491798389662]
  Previous works commonly design plug-and-play modules and structural losses to effectively extract and aggregate the global context. We propose Positive-Negative Equal contrastive loss (PNE loss), which increases the latent impact of positive embedding on the anchor and treats the positive as well as negative sample pairs equally. We conduct comprehensive experiments and achieve state-of-the-art performance on two benchmark datasets.
  arXiv  Detail & Related papers  (2022-07-04T13:51:29Z)
• Let Invariant Rationale Discovery Inspire Graph Contrastive Learning [98.10268114789775]
  We argue that a high-performing augmentation should preserve the salient semantics of anchor graphs regarding instance-discrimination. We propose a new framework, Rationale-aware Graph Contrastive Learning (RGCL) RGCL uses a rationale generator to reveal salient features about graph instance-discrimination as the rationale, and then creates rationale-aware views for contrastive learning.
  arXiv  Detail & Related papers  (2022-06-16T01:28:40Z)
• Chaos is a Ladder: A New Theoretical Understanding of Contrastive Learning via Augmentation Overlap [64.60460828425502]
  We propose a new guarantee on the downstream performance of contrastive learning. Our new theory hinges on the insight that the support of different intra-class samples will become more overlapped under aggressive data augmentations. We propose an unsupervised model selection metric ARC that aligns well with downstream accuracy.
  arXiv  Detail & Related papers  (2022-03-25T05:36:26Z)
• Learning One Representation to Optimize All Rewards [19.636676744015197]
  We introduce the forward-backward (FB) representation of the dynamics of a reward-free Markov decision process. It provides explicit near-optimal policies for any reward specified a posteriori. This is a step towards learning controllable agents in arbitrary black-box environments.
  arXiv  Detail & Related papers  (2021-03-14T15:00:08Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.