WeChat Researchers Propose WeGR, a Generative Retrieval Framework with Collapse-Free Tokenizer and RL
_reachsumit · x · 2026-10-14
A 12-author team published WeGR (arXiv:2610.16979), a unified generative retrieval framework for WeChat's industrial search that handles candidate generation within an autoregressive setup. Key components:
- CoCo-VAE tokenizer: injects query-to-item and item-to-item collaborative signals into Semantic IDs while mitigating codebook collapse via codebook refresh and complementary regularization.
- Dynamic Horizon Policy Optimization (DHPO): progressively expands reward evaluation and policy optimization horizons to combat reward sparsity under exact SID matching, exploiting behavioral rewards at different SID hierarchy levels.
The paper targets two long-standing bottlenecks in generative retrieval: collaborative-signal-aware tokenization and sparse RL rewards.
More from Research
- FrontierML benchmark tests whether agents can spend money to scale model performance — JayAlammar · 2026-10-14
- AIPOCH's Open-Science ships portable .science packages that share full research context, not just outputs — ahuja_priyank · 2026-10-14
- Video-Arena: Game-Based Video Benchmark Shows Native Models Score Just 13.2 F1 — Kuan Zhang · 2026-10-14
- LSCBench: Vibe Coding Agents Write Functional Code, But Only 25% Is Also Secure — Qwen · 2026-10-14
- Omni-LLMs Know More Than They Say: NAVER's OmniCatalyst Adds 7.29% via Self-Generated Text — naver-ai · 2026-10-14
- Repeated-Data Pretraining Study: Smaller Batches Act as Regularizers and Beat More GPUs — Sewon Min · 2026-10-14