2026-10-12 ~ 10-18 · Generative
Generative · Personalization · AI Search · Agentic · Infra
- 1. Whom Do AI Agents Work For? Role Assignment Induces Sponsorship Bias in LLM Recommenders
Changing who employs an assistant changes how it treats sponsored listings. Controlled hotel-choice experiments hold prices and reviews constant while varying delegation and sponsorship…
Appeared in: 10-16
- 2. Beyond ICD Codes: Semantic IDs for Generative Medical Coding
SemICD replaces raw diagnosis codes with hierarchical semantic identifiers. It clusters code descriptions and trains bidirectional description-to-ID tasks. Across four LLMs and two medical datasets…
Appeared in: 10-16
- 3. AutoSFT: An Agentic System for Automated Data Research in Supervised Fine-Tuning
AutoSFT gives agents a broader action space for training-data research: reweight, prune, repair, replace and synthesize. Execution, analysis and synthesis have separate roles…
Appeared in: 10-16
- 4. JevAgentHarness: Benchmarking, Deploying and Training Decision Models That Share Control with LLMs
JevAgentHarness delegates selected choices to a cheaper decision model and falls back to the LLM below calibrated thresholds. Training includes states reached after agent mistakes…
Appeared in: 10-16
- 5. ReaDy: Recurrent States as Dynamic Context for Speculative Drafting in Hybrid Linear Attention Large Language Models
ReaDy reads a hybrid linear-attention model's recurrent state to provide position-specific context for a speculative drafter. Low-rank reads and fused operators limit overhead…
Appeared in: 10-16
- 6. Trajectory-Level Speculation with Semantic Verification for Tool-Using Agents
SpecAgent lets a small model draft grounded tool trajectories and a strong model verify whole segments. Read-only calls execute during drafting; mutations wait for review. On τ²-Bench…
Appeared in: 10-16
- 7. TreeSpark: Learning to Draft the Trees You Verify for Speculative Decoding
TreeSpark trains a drafter on the candidate trees its verifier will actually inspect. It replays expansion, pruning and budgeted packing using existing target distributions…
Appeared in: 10-16
- 8. Online Tokenization with Optimal Delay in Linear Time
This work streams the tokenization pipeline instead of waiting for each stage to finish a complete input. It formalizes output delay and supplies theoretical bounds plus practical implementations…
Appeared in: 10-16
- 9. One Block Ahead: Promoting Sampler State in Diffusion Language Models
One Block Ahead denoises the next text block alongside the current one. It promotes tokens together with their sampler state, including denoising time and self-conditioning…
Appeared in: 10-16
- 10. ixi-sLM: On-Device sLM for AI Call Agent Service
ixi-sLM combines vocabulary trimming, a call-adapted tokenizer and task-aware mixed-precision quantization for mobile call assistance. One model powers call summarization and answering…
Appeared in: 10-16