Web This site Trusted
30 result(s) for “okapi bm25 site hotchips pp site does apache org site docs bell sw com site shs site:arxiv.org” of 35 matched · 510 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.
Did you mean okapi bm25 site hotchips pp site does apache org site docs bell sw com site she site:arxiv.org?
Freshness: newest result fetched 25 d ago · median age of these results 25 d · 1 of 30 carry a fetch time
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
6 Ingest Overhead: When Does Structured Memory Pay Off
https://arxiv.org/html/2608.03048v1
com/JetAstra/PI-Mem
Introduction
The ability to reason over long contexts is increasingly important for large language models (LLMs), supporting applications such as long-document question answering (Gemini Team 2025; Yang et al
https://arxiv.org/pdf/2608.02758
@۬MXT` )iI}= go)p 9I" "PZ AJBE&BZ@ae00)SV( )+ g H 0Jx cYfq;{~7zy:Zgޝi{c,c% Л*Ӂ`VNWwL/fW 2}(^tu0}efY ]3Co+iY/ڙg=nr\p֘EBp
https://arxiv.org/html/2609.02852v1
Description:
arXiv is now an independent nonprofit! Learn more ×
The Implications of Linguistic Illegibility for LLM Security
James Mickens Affiliation: Harvard University Email: mickens@g.harvard.edu
Abstract
LLMs are trained to
https://arxiv.org/pdf/2607.27188v1
pyFEˡF;{[W4rׯ_Rv+>SڂRv|u | = 韥~"'OuZwEGiDwX ),Pun&ߑ v" RX`> stream x[]s6}_Mg7$L7w&m4N hEӸ~IYvexh@pp H㴐8#GaEQ8LD酊e(Q0=Z{%= m yiQz ޏ"X
https://arxiv.org/src/2608.03883
lއxG&JWXԴ&[v @V[uԇF7vz%N` pP \>$kb {K/05 U0[|ʋX cc S-bQEAE
https://arxiv.org/src/2006.02833
p8Z fB*hOB8 Q6aF _qf @#ppk͓ҝI#}5J5CL
https://arxiv.org/src/2006.03271
{Ŗ 3*rTfϨ~{y>g +O:L'#i?KR} Uy3RoUyڃ ed8 w+%wV6XvN8W^;6ɄxUa Ke X$ j& Q8M _Pf٭;F/^S/̭ďI2Xc Cvt9
https://arxiv.org/src/2405.20232
(T$,J I>̙i9ɹ(Yd }6%! ?= `?9 tpr Z[~=(Jqp*U[ -ߴb9 6K}ڍ'wQˀqA ˦?5 JD/^Dͤȑ{V0z~w
https://arxiv.org/src/2006.01354
TdIjǖI,D ? =x`],Ă3֯v;!픻֭.t-W8I{參&-{s/}+Yp"X&v*i ; b"^ |aǶ >!ZQ}N?}=c߳? b? {.؏ '3'ѱ^[ͦa.ll X{,c7-ي}
https://arxiv.org/pdf/2407.00125
Reliability.Performance.- Software and its engineering -> Software organization and properties.- Computing methodologies -> Artificial intelligence.) /Title (A Survey on Failure Analysis and Fault Injection in AI Systems) /Trapped /False
https://arxiv.org/src/2006.01128
hLkol 'VV=kdgcfc``becf`zYYE.N vv{"ae"_ YbŧǷ3憢65s6gd`E'jalO'jml"lbdgl hb`徹-݆{WH_ lы>$1N RS$|E Dǫ`ݨǴE(M5nWAEQ16Xf qm
https://arxiv.org/src/2006.03594
6gb3=};9jB^MM=qL^pp> uNd: E n=ʢ(_o饖 X,(RcP9#4 ٜ+R0ʰǬe{'{";(: [ mjqU8y>erA( ;]V5Xclc}1eڊ06ȯ~Ps hGyqJ
https://arxiv.org/pdf/2005.13905
&Q,l@2`0!`01 $4g*S9g9~Pj?}7Z+7/m_^m7?GS_9Y,|?qAB JU$+ {aFB%5|!_F04 0k Ğt:fJ5^c`츾yZ IFv[Yb4 hE9''HC&3hj|w擒cM}iG{X
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/list/cs.CL/new
We evaluate five open-weight readers with Vanilla context, BM25-RAG, Oracle retrieval, Memobase, and MemSearch as memory backends
https://arxiv.org/pdf/2407.00345
obj >>> endobj 9 0 obj >>> endobj 10 0 obj >>> endobj 11 0 obj >>> endobj 12 0 obj >>> endobj 13 0 obj > stream xZێ7 }Ws("E]v{/t5 Nu樂Vbel%, 48 >D)PtPaoёK{k~bS+"X` +b
https://arxiv.org/html/2407.16741v3
com/All-Hands-AI/OpenHands
Slack http://bit
https://arxiv.org/html/2607.26327v3
8 What the simulation does not show
6 Discussion
6
https://arxiv.org/src/2212.07841
s sWd2u ]۳M fzz `bFLHC} mw"3W֓ nA\ q@}[yi( ) U]Czj/ntUm VZ+bTzf͓4L7m:_9 9MG$
https://arxiv.org/src/1811.12628
SQ6) Sν3{ǺN{|rϽe,*s5 dee?'O{v==#+##3#-3͞fg6blA $"4ANԮ?)d-8* ,e W⼒_b| -1lǔqsDyew0usAF?QW2#V{5/W*wen:y\ae \' ЃgE !
https://arxiv.org/list/cs.AI/new
Research artifact available under the Apache-2
https://arxiv.org/src/2608.03930
pw\W[pY Dsw"%EƳ vgG\ U -u G\Wt,`s·|y'']U&kܡ]d/ga 1X Tq=x&1|/d* Y4HS|]>ԘX
https://arxiv.org/pdf/2607.28577v1
{ťu$]pp8C2N҂BYARxN2JNr3N )I1B$H) `(7IʒP02 a*pT J8.9iJ(="@ a^f◷o=X"? 1_Q쫶"|{7JQ/VJ>H?t'XorG )g+Wh:g?O1 Zn
https://arxiv.org/html/2608.02999v1
3 Dictionary coverage does not identify a reading
7 Discussion
What the construction establishes
https://arxiv.org/pdf/2212.07841
%*fcXPp@A|bd"8hnb
https://arxiv.org/pdf/2608.02615
YoRDKhr4 Y-["4}H d@\Dȃ(@IJ 88ࠐiAAM ոD I 9`4ki( * *%@xyK]^ $5۫nJzoQm6aHMf;la6gXڣ=xN ,W]ٌ x
https://arxiv.org/list/cs.IR/new
Skip to main content
System maintenance August 4th and 5th Learn more ×
Search Submit Donate Log in
Search arXiv
Press Enter to search · Advanced search
Information Retrieval
New submissions
Cross-lists
Replacements
See recent
https://arxiv.org/abs/2608.03883
Our systems combine sparse BM25 retrieval, evidence-aware rescoring for self-referential symptom reports, embedding-based reranking, query-prototype expansion, and LLM-based reranking
https://arxiv.org/abs/2104.08663
Our results show BM25 is a robust baseline and re-ranking and late-interaction-based models on average achieve the best zero-shot performances, however, at high computational costs