Beyond Post-Hoc Rewards: Privileged Reasoning Exploration for Multimodal Embeddings

A new framework called PILOT, described in arXiv paper 2610.32474v1, improves multimodal embedding models by treating reasoning as a latent variable and co-evolving a prior mode conditioned only on the input with a posterior mode that also uses training-time relational evidence. On the 78-task MMEB-V2 benchmark, PILOT beats the primary baseline across all 12 task groups at both 2B and 7B parameters, raising overall scores by 3.2 and 3.3 points, with the largest gains in visual-document retrieval (+6.1 and +6.4 points). The authors conclude that relational supervision is most valuable as a signal for shaping how reasoning is explored during training rather than as a post-hoc reward.

arXiv:2610.32474v1 Announce Type: cross Abstract: Advances in Multimodal Large Language Models MLLMs have driven substantial progress in multimodal embeddings. Recent methods exploit generative capabilities by augmenting embedded inputs with self-generated reasoning to produce more informative embeddings. However, conventional retrieval supervision for training embedding models specifies pairwise relevance relations without explicitly guiding the reasoning used for input augmentation. Such reasoning is neither directly observed nor uniquely determined by the input, allowing multiple plausible traces to introduce semantic cues of varying retrieval utility. Existing approaches largely optimize reasoning conditioned on the input alone, limiting the exploration of retrieval-relevant traces. We introduce PILOT, a framework that models reasoning as a latent variable and instantiates its prior and posterior distributions as distinct reasoning modes of a shared MLLM. The prior conditions only on the input, whereas the posterior uses additional training-time relational evidence to guide reasoning exploration. To translate the posterior's broader exploration into inference-time improvements, we derive an importance-weighted variational objective and instantiate it as a prior-posterior co-evolution framework that improves both modes under a shared inference-compatible reward while selectively transferring posterior trajectories to the prior according to their utility and prior reachability. On the 78-task MMEB-V2 benchmark, PILOT outperforms the primary baseline across all 12 task groups at both 2B and 7B parameters, improving overall scores by 3.2 and 3.3 points, respectively, with the largest gains in visual-document retrieval +6.1 and +6.4 points . Our results show that relational supervision is most valuable not as a post-hoc reward but as a signal for shaping how reasoning is explored in training reasoning-enhanced embedding models.