Dzokera kuNhau
InnovationAI Understanding muchidimbu

Vatsvagiri vanopa AutoTuneBench yekuyero yakavimbika yeLLM inoshumira injini

Chiyereso chitsva chekuyera uye kuyerwa kweprotocol yevamiririri vemienzaniso yemutauro mikuru inoratidzwa, ichigadzirisa nzira ina dzekutadza munzira dziripo dzekuyera.

By 4 min readRead the primary source
Source-page capture accompanying Researchers present AutoTuneBench for trustworthy measurement of LLM serving engines
Primary-source documentKwakanyorwa
Muparidzi
arxiv.org
Source link
arxiv.orghttps://arxiv.org/abs/2609.18123
Source type
Gwaro rekutanga - chiziviso chepamutemo, bepa, faira, kana peji rebato rekutanga ratinoverenga zvakananga.
ContextNzwisisa izvi mumasekonzi makumi matanhatu

Tanga pano

Matemu akakosha

Mutauro Mukuru (LLM)
Mutauro wemodhi yakadzidziswa pane yakakura text corpora kugadzira nekuongorora zvinyorwa.
Benchmark
Muedzo wakamisikidzwa kana dhatabheti rinoshandiswa kuyera nekuenzanisa kuita kwemuenzaniso.
Zviedze iwe pachakoChii chinonzi AI? Quiz

Chii chaitika

Vatsvagiri vakagadzira AutoTuneBench, bhenji uye kuyerwa kweprotocol yevamiriri vemitauro mikuru. Iyo protocol inogadzirisa nzira ina dzekutadza munzira dziripo dzekuyera: strawman baselines, isingachinjike zvachose nguva, yakazara mabasa, uye hurema hwezvivakwa. AutoTuneBench yakagadzirirwa kupa zviyero zvinovimbika nekuomesa iyo protocol sekodhi, kusimbisa test-provenance, uye kushandisa database-level validator kuramba kunze-kwe-protocol mhinduro.

Vatsvagiri vakaratidza nzira ina dzekutadza munzira dziripo dzekuyera: strawman baselines, isingachinjike zvachose nguva, mabasa akazara, uye hurema hwezvivakwa.

AutoTuneBench inogadzirisa nzira idzi dzekukundikana nekuomesa protocol sekodhi, kusimbisa test-provenance, uye kushandisa database-level validator kuramba kunze-kwe-protocol mhinduro.

Iyo protocol yakagadzirirwa kupa zviyero zvinovimbika nekusimbisa kune zvakaburitswa kunze mhedzisiro, yakavakirwa pambiri-mhodzi nhamba ine 5% cross-run coefficient-of-variation cap.

Vatsvakurudzi vakaratidza kushanda kweAutoTuneBench nekuenzanisa migumisiro yayo kune iyo yemaitiro ekuyera aripo.

Mhedzisiro yacho yakaratidza kuti AutoTuneBench yakapa zviyero zvakaringana uye zvakavimbika, nekumhanya kwepakati kwe1.0001x pamusoro pePyTorch ichida.

Kwakabva mashoko: arxiv.org ↗

Nei zvichikosha

Kuvandudzwa kweAutoTuneBench kwakakosha nekuti inogadzirisa nyaya yakakosha mumunda wevamiriri vemitauro mikuru. Nzira dzekuyera dziripo dzinowanzo kanganisa, zvichiita kuti pave nemigumisiro isina kururama uye isina kuvimbika. AutoTuneBench inopa yakavimbika kuyerwa protocol iyo inogona kushandiswa kuongorora mashandiro eLLM vamiririri. Izvi zvinonyanya kukosha kune zvikumbiro apo kuyerwa kwakarurama uye kwakavimbika kwakakosha, senge mukuita sarudzo dzepamusoro-soro kana mumamiriro ezvinhu apo mhedzisiro yezviyero zvisiri izvo zvakanyanya.

Kuvandudzwa kweAutoTuneBench kwakakosha nekuti inogadzirisa nyaya yakakosha mumunda wevamiriri vemitauro mikuru.

Nzira dzekuyera dziripo dzinowanzo kanganisa, zvichiita kuti pave nemigumisiro isina kururama uye isina kuvimbika.

AutoTuneBench inopa yakavimbika kuyerwa protocol iyo inogona kushandiswa kuongorora mashandiro eLLM vamiririri.

Izvi zvinonyanya kukosha kune zvikumbiro apo kuyerwa kwakarurama uye kwakavimbika kwakakosha, senge mukuita sarudzo dzepamusoro-soro kana mumamiriro ezvinhu apo mhedzisiro yezviyero zvisiri izvo zvakanyanya.

Kuvandudzwa kweAutoTuneBench kunosimbisa kudiwa kwenzira dzakaomarara uye dzakavimbika dzekuyera mumunda weAI.

Interactive Mechanism

Interactive Mechanism: Iyo Inonyatsoshanda

Ongorora ari pasi tekinoroji kuseri kwekusimudzira uku uchipindirana.

Agent Lifecycle Stage:
1
User Intent & Planning: "Audit customer refund request #4092 and settle payment."
2
Tool Calling: Emits structured JSON call crm_get_transaction(id='4092').
3
Guardrail & Verification:🛡️ Paused: High-value action requires human operator sign-off.
4
Final Settlement: Refund recorded, email receipt dispatched, and audit log stored.
Core takeaway: An AI agent is not just a language model—it is a closed loop of planning, tool invocation, and environment feedback. Production systems require self-healing retries and strict human approval guardrails.
Interactive Concept Check+10 Points
What is AI? Quiz

A route planner searches possible journeys using explicit rules. What does this illustrate about AI?

Zvekutarisa zvinotevera

Mhedzisiro yeAutoTuneBench pandima yevamiririri vemitauro mikuru ichave yakakosha. Ichapa yakavimbika kuyerwa protocol iyo inogona kushandiswa kuongorora mashandiro eLLM vamiririri. Izvi zvinogonesa vaongorori nevagadziri kuti vatore zviyero zvakaringana uye zvakavimbika, zvichitungamira mukuita sarudzo kuri nani uye mamwe maapplication anoshanda. Pamusoro pezvo, kuvandudzwa kweAutoTuneBench kunosimbisa kudiwa kwenzira dzakaomarara uye dzakavimbika dzekuyera mumunda weAI.

Mhedzisiro yeAutoTuneBench pandima yevamiririri vemitauro mikuru ichave yakakosha.

Ichapa yakavimbika kuyerwa protocol iyo inogona kushandiswa kuongorora mashandiro eLLM vamiririri.

Izvi zvinogonesa vaongorori nevagadziri kuti vatore zviyero zvakaringana uye zvakavimbika, zvichitungamira mukuita sarudzo kuri nani uye mamwe maapplication anoshanda.

Pamusoro pezvo, kuvandudzwa kweAutoTuneBench kunosimbisa kudiwa kwenzira dzakaomarara uye dzakavimbika dzekuyera mumunda weAI.

Vatsvagiri vanoronga kuenderera mberi nekuvandudza uye kuchenesa AutoTuneBench kuti ive nechokwadi chekushanda kwayo uye kuvimbika.

Related guides & Quizzes

Chii chinonzi AI?AI AgentsAI Models InotsanangurwaTransformersKudzidziswa kweAIEdza zvaunoziva - edza yemahara AI quizTarisa kumusoro izwi reAI mune yedu glossaryTevedza iyo AI modhi yekuburitsa tracker
Wakawana izvi zvinobatsira?