AI Search
When rewriting a query vector with first-round retrieved documents or an LLM-written hypothetical answer (HyDE), the usual approach averages feedback vectors and interpolates with the original query.
Personalization
Taobao search CTR models mainly remember historical clicks through item IDs, so ==the same item with a different main image is almost indistinguishable to the model==.
Personalization
Generative recommenders use beam search to emit item-ID tokens one by one, and ==tail items have low generation probability and are often pruned in the first one or two steps==: on Amazon Industrial & Scientific, MiniOneRec's 50 candidates cover only 11.94% of tail targets.
Personalization
Personalized reward models score replies per user, but each user labels only a few dozen preference pairs, so most unlabeled replies go unused.
Personalization
When a sequential recommender encodes each historical behavior, ==some behaviors converge after one layer while others need several stacked layers to stabilize==.
Infra
An independent researcher compares "add 2:4 sparsity" versus "drop one precision tier" on Llama-3.1-8B-Instruct and Qwen3-8B under matched storage bits (including sparse indices) and matched nominal compute tiers, with measurements on L40S, H100, and RTX PRO 6000. ==As long as a lower precision is still available, adding 2:4 is both less accurate and slower==: 2:4 + INT8 costs the same storage as INT5.
Infra
SGLang, NVIDIA, and AMD use a plugin framework to plug specialized fused kernels into SGLang, cutting inter-kernel waits and cross-GPU communication at low concurrency.
Infra
Hisense's RCSD continues training a speculative-decoding draft model that was distilled with HASS, using DPO or GRPO: sequences from the target alone are "good," sequences alternating target and draft are "bad," and the GRPO variant scores by acceptance rate.
Infra
Scitix.AI's UlaCache maps every GPU KV pool and CPU memory into one logical address space so ==GPU and CPU each store different history==, and removes MLA's cross-GPU duplication under TP; the GPU pool is shared across requests with a bucketed parallel LRU for eviction.
Generative
Generative recommenders emit item-ID tokens left to right, so temporary diversity at inference ==can only rewrite one position mid-decode and often produces IDs that do not exist in the catalog==.
Agentic
Korea University and KAIST have Claude Code (Claude Opus 5.5) design retrieval tools for a search agent: it reads each question's execution log and score, jointly edits pre-generated index files, executable search scripts, and usage instructions, and after 10 rounds keeps the best version on the optimization set.
Agentic
Northeastern University and Tsinghua find that in retrieve-while-reasoning RAG, ==later sub-queries keep drifting away from the original question along the direction earlier sub-queries already moved==, and the effect grows with step count.
Agentic
Parlant used to ask an LLM whether each customer-service rule is relevant every turn—about 69,000 tokens for 36 rules.
Agentic
IBM builds ClariEdge, a ~3,900-sample multi-turn RAG clarification set that specially includes hard negatives that "look ambiguous but the referent is already fixed by context," then LoRA-tunes small Granite models to decide whether to clarify and to list candidates.