Web This site Trusted
30 result(s) for “okapi bm25 site hotchips pp site github com site text of de site may director site:arxiv.org” of 37 matched · 1088 ms
Scope: Nishi’s own index (our crawl + Common Crawl) — not the whole live web. Some sites block all automated crawlers (anti-bot walls), so their pages may be absent here even though we never filter lawful content.
Freshness: no result on this page carries a fetch time yet (indexed before the fetch epoch was stored beside the doc)
The query asks the model to recall the topic of an earlier conversation about Tesla; the gold answer is anchored in one specific session about Pooja’s dig site
from the top result below
https://arxiv.org/html/2608.02613v1
(3) Search latency bites only at very small reader: on a Spark GB10 edge node, memory-search adds a moderate and fixed 8787/77/4848 ms (BM25-RAG/Memobase/MemSearch) that composes a small part of TTFT for most reader-backend combinations
https://arxiv.org/html/2608.03048v1
To address these issues, we propose PI-Mem (Parallel-Iterative Memory), a mechanism that processes all chunks in parallel and iteratively refines a shared memory over a bounded number of turns
https://arxiv.org/html/2608.03063v1
Merchant risk control at large payment platforms screens tens of millions of merchants per day, where errors are costly in both directions: a false positive may harm a legitimate merchant, while a false negative leaves harmful activity undetected
https://arxiv.org/list/cs.AI/new
However, their performance heavily depends on the clarity and completeness of data semantics
https://arxiv.org/html/2608.03011v1
AS] 04 Aug 2026
Towards Real-world Environment-aware Zero-shot Text-to-speech Synthesis via Disentangled Audio Infilling
Ye-Xin Lu, Xin Wang, Yang Ai, Hui-Peng Du, Zhen-Hua Ling, and Junichi Yamagishi This work is the extended version of our conference paper [1] published at IEEE International Conference on
https://arxiv.org/html/2501.13956v1
com
Abstract
We introduce Zep, a novel memory layer service for AI agents that outperforms the current state-of-the-art system, MemGPT, in the Deep Memory Retrieval (DMR) benchmark
https://arxiv.org/html/2608.02668v1
Both prove to be members of a one-parameter family of angular retractions, pp-SpheretNorm, whose rotation angle saturates rather than growing without bound
https://arxiv.org/html/2509.22186v2
ocr, PP-StructureV3), establishing new performance records in text recognition, formula recognition, table recognition, and reading order prediction
https://arxiv.org/html/2407.04503v4
team, Université de Bordeaux, France
Abstract
As large language models (LLMs) start interacting with each other and generating an increasing amount of text online, it becomes crucial to better understand how information is transformed as it passes from one LLM to the next
https://arxiv.org/html/2407.16741v3
2 Agent Runtime: How Execution of Actions Results in Observations
2
https://arxiv.org/list/cs.CL/new
(3) Search latency bites only at very small reader: on a Spark GB10 edge node, memory-search adds a moderate and fixed 87/7/48 ms (BM25-RAG/Memobase/MemSearch) that composes a small part of TTFT for most reader-backend combinations
https://arxiv.org/html/2608.02918v1
Biscainho are with SMT, DEL/Poli & PEE/COPPE, Federal University of Rio de Janeiro, Brazil (corresponding e-mail: wallace
https://arxiv.org/html/2608.03486v1
Much of the literature has focused on linear specifications, where the reduced-form innovations are assumed to be linear combinations of the structural shocks and identification may be achieved, for example, through exclusion restrictions, external instruments, heteroskedasticity, or non-Gaussianity (see, e
https://arxiv.org/html/2607.22511v2
A statement can be free of sorry and still be mathematically uninformative: a definition may reduce to True; or an unproven step may be introduced as an axiom that the kernel accepts
https://arxiv.org/html/2608.02616v1
2 Related Work
De-identification and text anonymization
https://arxiv.org/html/2608.02609v1
Code and pretrained model: https://github
https://arxiv.org/html/2210.00131v4
For example, underspecification in Figure 1b may serve to amplify its gender-occupation shortcut bias relative to that of Figure 1a
https://arxiv.org/html/2608.02612v1
2 Behavioral Analysis of LLMs as Meta-Solvers
Sensitivity to Problem Domains and Difficulties
https://arxiv.org/html/2407.02518v2
For instance, in a study of Github’s Copilot, Pearce et al
https://arxiv.org/html/2608.03372v1
The check compares the stored text against its source; the question is what the comparison is made of: a free, auditable word list, or an LLM call that must be trusted
https://arxiv.org/html/2508.19205v1
In our experiments, this highly efficient tokenizer maintains a speech-to-text token ratio of approximately 2:1, meaning two speech tokens are roughly equivalent to one BPE SHB (15) text token
https://arxiv.org/html/2608.03446v1
Predicting Multilingual Classification and Translation Performance of LLMs with Cross-Lingual Alignment – Is English Enough
https://arxiv.org/html/2407.17789v2
Assigning agents with simple backgrounds may result in a significant number of highly homogeneous agents, making it difficult to derive meaningful insights
https://arxiv.org/abs/2608.03883
The task requires systems to rank candidate Reddit sentences according to their relevance to each of the 18 symptoms in the Adult ADHD Self-Report Scale (ASRS-v1
https://arxiv.org/abs/2104.08663
We leverage a careful selection of 18 publicly available datasets from diverse text retrieval tasks and domains and evaluate 10 state-of-the-art retrieval systems including lexical, sparse, dense, late-interaction and re-ranking architectures on the BEIR benchmark
https://arxiv.org/abs/2608.03860
The pipeline combines sentence-window chunking, BM25, BGE-M3 dense retrieval, reciprocal rank fusion, optional cross-encoder reranking, and grounded answer generation
https://arxiv.org/list/cs.IR/new
For what to learn, conventional next-token prediction treats adjacency as dependency and may encode spurious transitions across unrelated sessions
https://arxiv.org/html/2506.01844v1
The success of text-based models has thus been extended to other modalities, sparking interest towards multi-modal vision-language (VLMs) (Alayrac et al
https://arxiv.org/html/2608.00785v1
com
Abstract
Content discovery systems depend on relevance judgment for search quality evaluation, but human annotation faces inter-annotator disagreement and scaling costs
https://arxiv.org/html/2508.16279v1
We provide a scalable evaluation module with a visual studio interface, making the development of long-trajectory agentic applications more manageable and easier to trace