プロセス制御x 自然言語処理の現在地と展望 / Current State and Future Outlook for Process Control and Natural Language Processing

2026年1月23日「AI制御ワーキング第１回研究会」発表資料
URL: https://www.psec.jp/ai%E5%88%B6%E5%BE%A1wg%E7%AC%AC1%E5%9B%9E%E7%A0%94%E7%A9%B6%E4%BC%9A/

Transcript

1. ϓϩηε੍ޚ Y
   ࣗવݴޠॲཧͷݱࡏ஍ͱల๬ Ճ౻ ↅଠ TIPUB!IVNBOTZTJLZPUPVBDKQ ژ౎େֶେֶӃ ৘ใֶݚڀՊ ώϡʔϚϯγεςϜ࿦෼໺ 2025-01-23 AI制御ワーキング第１回研究会
2. ࣗݾ঺հɿՃ౻
   ↅଠ 1 
   ژେӃ޻ֶݚڀՊ Խֶ޻ֶઐ߈मྃ 
   ಉେֶӃ৘ใֶݚڀՊ γεςϜՊֶઐ߈मྃ r ಉॿڭ r
   ϚϯνΣελʔେֶ ٬һݚڀһ r ࢈૯ݚ
   টᡈݚڀһ ઐ໳ɿԽֶ޻ֶ
   Y
   σʔλαΠΤϯε
   Y
   ࣗવݴޠॲཧ ݚڀ಺༰ • ੡଄ϓϩηεͷϞσϦϯάɺ੍ޚ • ෺ཧϞσϧߏஙͷޮ཰Խ 9
   5XJUUFS 
   !T@LBU झຯ
   ΧϝϥɼαΠΫϦϯάɼΰϧϑɼ ΞΠϦογϡ΢ΠεΩʔʢ#VTINJMMTʣɼήʔϜ
3. ࠷ۙͷେن໛ݴޠϞσϧؔ࿈χϡʔε 2 https://news.yahoo.co.jp/pickup/6566805 https://support.claude.com/ja/articles/13345190 https://youtu.be/WBNZpAWhw5E?si=p4yVraQJqUuzfzi_
4. • ݴޠϞσϧʢ-.ʣ͸୯ޠྻͷ֬཰
   𝑝(𝑥! , ⋯ , 𝑥" ) Λ༧ଌ͢Δɽ • -.Λ༻͍Δ͜ͱͰɼ୯ޠྻʹଓ͘୯ޠΛ༧ଌͰ͖Δɽ • େن໛ݴޠϞσϧͱ͸ɼେن໛ͳσʔλͰֶश͞Εͨύϥϝʔλ਺ͷଟ͍-.ɽ -BSHF
   MBOHVBHF
   NPEFM
   --. • جຊతʹ͸ੑೳ㲍ύϥϝʔλ਺ɽ࠷ۙ͸#Λ௒͑ΔPQFOXFJHIU
   --.΋૿͖͑ͯͨɽ ݴޠϞσϧʢ-BOHVBHF
   .PEFM
   -.ʣ 3 𝑝 京都, の, 名物, は, 納豆 = 0.00000082 𝑝 京都, の, 名物, は, ラーメン = 0.00000197 𝑝 京都, の, 名物, は, 寿司 = 0.00000019 ɾɾɾ ژ౎ͷ໊෺͸ʁʁʁ / 𝑥" = argmax 𝑝 𝑥! , ⋯ , 𝑥"#! , 𝑥" 𝑥" ʙԯʢ#ʣݸ
5. طଘͷLLM 4 [Zhao+, arXiv, 2025] 2024. 5 GPT-4o GPT-3.5
6. • ೖྗʢϓϩϯϓτʣΛ޻෉͢Δ͚ͩͰ༷ʑͳλεΫΛղ͚Δɽ • େن໛ݴޠϞσϧͰߴ͍ੑೳΛୡ੒͢ΔͨΊʹɼ ଟ͘ͷϓϩϯϓτઃܭํ๏͕ఏҊ͞Ε͍ͯΔ [Liu+, ACM Comput. Surv., 2023]ɽ • จ຺಺ֶशʢJODPOUFYU
   MFBSOJOH
   *$-ʣ ʢGFXTIPU
   ֶशͱ΋ݺͿʣ[Brown+, NeurIPS, 2020] • ࢥߟ࿈࠯ܕϓϩϯϓτ $IBJOPG5IPVHIU
   QSPNQUJOH
   $P5
   [Wei+, NeurIPS, 2022] • 4FMG
   DPOTJTUFODZʢଟ਺ܾʣ [Wang+, ICLR, 2023] • LLMʹΑΔߴੑೳୡ੒ͷͨΊʹɼ Ϟσϧʢֶशσʔλ΍ύϥϝʔλ਺ɼreasoningͷ༗ແͳͲʣͷબఆɼ λεΫʹಛԽͨ͠ϫʔΫϑϩʔͷఆٛͳͲ͕ߦΘΕͯΔɽ --.͸Կ͕Ͱ͖Δ͔ʁ 5
7. • 5IF
   "*
   4DJFOUJTU
   --.ʹΑΔΞΠσΞੜ੒ɾ࣮ݧɾ࿦จࣥචͷࣗಈԽ • 1ZUIPOίʔυ΍࣮ݧ༻σʔληοτͰߏ੒͞ΕΔʮςϯϓϨʔτʯΛվળ͍ͯ͘͠ɽ • --.ͷ࿦ཧੑͱਖ਼֬ੑΛ޲্ͤ͞ΔͨΊʹɼ༷ʑͳϓϩϯϓτઃܭํ๏Λ׆༻͍ͯ͠Δɽ λεΫʹಛԽͨ͠ϫʔΫϑϩʔ
   ͷྫɿ5IF
   "*
   4DJFOUJTU 6 [Lu+, arXiv, 2024]
8. • --.ʹɼ࣮૷໨తΛೖྗͨ͠Β࣮૷ίʔυͱ݁Ռ͕ಘΒΕΔɽ ͕ɼΑ͘໰୊͕ى͜Δɽ • --.
   ͕ೖྗͷઐ໳༻ޠΛཧղͰ͖ͳ͍ɽ • --.
   ͷग़ྗܗ͕ࣜࢦࣔ௨Γͱ͸ҟͳΔɽ • ίʔυ͸ಘΒΕ͕ͨɼಈ͔ͳ͍ɽ • ҙਤ௨Γͷ࣮૷Λͤͣɼྫ֎ॲཧͷΈΛ࣮૷͢Δɽ • ʜ ྫɿ--.ʹΑΔ৽੍ޚख๏ͷ࣮૷ίʔσΟϯά 7 ʻೖྗʼख๏"ΑΓ΋ߴੑೳͳख๏ ΛఏҊ͢Δɽର৅ܥ͸ɼϓϩηε9 ͱ͠ɼγϛϡϨʔγϣϯͰ༗༻ੑ Λࣔ͢ɽ --. ʻग़ྗʼϓϩηε9Λର৅ͱͨ͠ ࣮ݧʹ͓͍ͯɼఏҊख๏͕ख๏" ΑΓ΋༏Ε͍ͯΔ͜ͱΛ࣮ূ͠· ͨ͠ɽίʔυ͸ɼʜ
9. • --.ʹɼ࣮૷໨తΛೖྗͨ͠Β࣮૷ίʔυͱ݁Ռ͕ಘΒΕΔɽ ͕ɼΑ͘໰୊͕ى͜Δɽ • --.
   ͕ೖྗͷઐ໳༻ޠΛཧղͰ͖ͳ͍ɽˠ
   3"(
   ʹΑΔ஌֦ࣝॆ • --.
   ͷग़ྗܗ͕ࣜࢦࣔ௨Γͱ͸ҟͳΔɽ • ίʔυ͸ಘΒΕ͕ͨɼಈ͔ͳ͍ɽ • ҙਤ௨Γͷ࣮૷Λͤͣɼྫ֎ॲཧͷΈΛ࣮૷͢Δɽ • ʜ ྫɿ--.ʹΑΔ৽੍ޚख๏ͷ࣮૷ίʔσΟϯά 8 ʻೖྗʼख๏"ΑΓ΋ߴੑೳͳख๏ ΛఏҊ͢Δɽର৅ܥ͸ɼϓϩηε9 ͱ͠ɼγϛϡϨʔγϣϯͰ༗༻ੑ Λࣔ͢ɽ --. ʻग़ྗʼϓϩηε9Λର৅ͱͨ͠ ࣮ݧʹ͓͍ͯɼఏҊख๏͕ख๏" ΑΓ΋༏Ε͍ͯΔ͜ͱΛ࣮ূ͠· ͨ͠ɽίʔυ͸ɼʜ
10. • --.
    ʹ଍Γͳ͍஌ࣝΛจॻ͔Βऩू͠ɼϓϩϯϓτʹؚΊΔɽ • ΫΤϦϕΫτϧͱࣅͨຒΊࠐΈϕΫτϧΛ࣋ͭจॻΛؔ࿈จॻͱͯ͠औಘ͢Δɽ • ղ͖͍ͨ໰୊ʹର͢Δσʔληοτɾద੾ͳຒΊࠐΈΛੜ੒͢Δख๏͕ඞཁɽ ݕࡧ֦ுੜ੒ʢ3FUSJFWBM"VHNFOUFE
    (FOFSBUJPOʣ 9 [Lewis+, NeurIPS, 2020]
11. • --.ʹɼ࣮૷໨తΛೖྗͨ͠Β࣮૷ίʔυͱ݁Ռ͕ಘΒΕΔɽ ͕ɼΑ͘໰୊͕ى͜Δɽ • --.
    ͕ೖྗͷઐ໳༻ޠΛཧղͰ͖ͳ͍ɽˠ
    3"(
    ʹΑΔ஌֦ࣝॆ • --.
    ͷग़ྗܗ͕ࣜࢦࣔ௨Γͱ͸ҟͳΔɽˠ
    ߏ଄Խग़ྗʹΑΔܗ੍ࣜݶ • ίʔυ͸ಘΒΕ͕ͨɼಈ͔ͳ͍ɽ • ҙਤ௨Γͷ࣮૷Λͤͣɼྫ֎ॲཧͷΈΛ࣮૷͢Δɽ • ʜ ྫɿ--.ʹΑΔ৽੍ޚख๏ͷ࣮૷ίʔσΟϯά 10 ʻೖྗʼख๏"ΑΓ΋ߴੑೳͳख๏ ΛఏҊ͢Δɽର৅ܥ͸ɼϓϩηε9 ͱ͠ɼγϛϡϨʔγϣϯͰ༗༻ੑ Λࣔ͢ɽ --. ʻग़ྗʼϓϩηε9Λର৅ͱͨ͠ ࣮ݧʹ͓͍ͯɼఏҊख๏͕ख๏" ΑΓ΋༏Ε͍ͯΔ͜ͱΛ࣮ূ͠· ͨ͠ɽίʔυ͸ɼʜ
12. • --.
    ͷग़ྗܗࣜΛݻఆ͢Δɽ • ࣗ༝จ͸ݕূ͕ࠔ೉ͰͰ͋ΔͨΊɼεΩʔϚΛఆٛ͢Δɽ • ߏ଄ͱ಺༰ΛΘ͚ͯධՁ͢Δ͜ͱͰอकɾվળ͠΍͘͢͢Δɽ • +40/ܗࣜͷग़ྗ͕Α͘༻͍ΒΕ͍ͯΔɽ [Geng+, arXiv, 2025] • ߏ଄Խ͢Δ͜ͱͰɼطଘͷϧʔϧνΣοΫ ΋׆༻Ͱ͖Δɽ ݕূՄೳʹ͢ΔͨΊͷඞਢٕज़ɿߏ଄Խग़ྗ 11 ߏ଄Խͳ͠ ʻೖྗʼख๏"ΑΓ΋ߴੑೳͳख๏ ΛఏҊ͢Δɽର৅ܥ͸ɼϓϩηε9 ͱ͠ɼγϛϡϨʔγϣϯͰ༗༻ੑ Λࣔ͢ɽ --. ʻग़ྗʼϓϩηε9Λର৅ͱͨ͠ ࣮ݧʹ͓͍ͯɼఏҊख๏͕ख๏" ΑΓ΋༏Ε͍ͯΔ͜ͱΛ࣮ূ͠· ͨ͠ɽίʔυ͸ɼʜ ߏ଄Խ͋Γ ର৅ϓϩηεɿϓϩηε9 ม਺ɿ<7
    U
    ʜ> طଘख๏ʹΑΔੑೳɿʜ ʜ
13. • --.ʹɼ࣮૷໨తΛೖྗͨ͠Β࣮૷ίʔυͱ݁Ռ͕ಘΒΕΔɽ ͕ɼΑ͘໰୊͕ى͜Δɽ • --.
    ͕ೖྗͷઐ໳༻ޠΛཧղͰ͖ͳ͍ɽˠ
    3"(
    ʹΑΔ஌֦ࣝॆ • --.
    ͷग़ྗܗ͕ࣜࢦࣔ௨Γͱ͸ҟͳΔɽˠ
    ߏ଄Խग़ྗʹΑΔܗ੍ࣜݶ • ίʔυ͸ಘΒΕ͕ͨɼಈ͔ͳ͍ɽˠ
    --.ʹΑΔπʔϧ࢖༻ • ҙਤ௨Γͷ࣮૷Λͤͣɼྫ֎ॲཧͷΈΛ࣮૷͢Δɽ • ʜ ྫɿ--.ʹΑΔ৽੍ޚख๏ͷ࣮૷ίʔσΟϯά 12 ʻೖྗʼख๏"ΑΓ΋ߴੑೳͳख๏ ΛఏҊ͢Δɽର৅ܥ͸ɼϓϩηε9 ͱ͠ɼγϛϡϨʔγϣϯͰ༗༻ੑ Λࣔ͢ɽ --. ʻग़ྗʼϓϩηε9Λର৅ͱͨ͠ ࣮ݧʹ͓͍ͯɼఏҊख๏͕ख๏" ΑΓ΋༏Ε͍ͯΔ͜ͱΛ࣮ূ͠· ͨ͠ɽίʔυ͸ɼʜ
14. • --.
    ͕֎෦πʔϧʢQZUIPO
    ݕࡧػೳͳͲʣΛ࢖͍ॲཧΛ࣮ߦ͢Δɽ • πʔϧΛʮ͍ͭʯʮԿΛʯʮͲͷҾ਺Ͱʯߦ͏͔Λఆٛ͢Δɽ ֎෦πʔϧͷMར༻ 13 [Guo+, ACL, 2024]
15. • --.ʹɼ࣮૷໨తΛೖྗͨ͠Β࣮૷ίʔυͱ݁Ռ͕ಘΒΕΔɽ ͕ɼΑ͘໰୊͕ى͜Δɽ • --.
    ͕ೖྗͷઐ໳༻ޠΛཧղͰ͖ͳ͍ɽˠ
    3"(
    ʹΑΔ஌֦ࣝॆ • --.
    ͷग़ྗܗ͕ࣜࢦࣔ௨Γͱ͸ҟͳΔɽˠ
    ߏ଄Խग़ྗʹΑΔܗ੍ࣜݶ • ίʔυ͸ಘΒΕ͕ͨɼಈ͔ͳ͍ɽˠ
    --.ʹΑΔπʔϧ࢖༻ • ҙਤ௨Γͷ࣮૷Λͤͣɼྫ֎ॲཧͷΈΛ࣮૷͢Δɽˠ
    ΤʔδΣϯτʹΑΔ֬ೝ • ʜ ྫɿ--.ʹΑΔ৽੍ޚख๏ͷ࣮૷ίʔσΟϯά 14 ʻೖྗʼख๏"ΑΓ΋ߴੑೳͳख๏ ΛఏҊ͢Δɽର৅ܥ͸ɼϓϩηε9 ͱ͠ɼγϛϡϨʔγϣϯͰ༗༻ੑ Λࣔ͢ɽ --. ʻग़ྗʼϓϩηε9Λର৅ͱͨ͠ ࣮ݧʹ͓͍ͯɼఏҊख๏͕ख๏" ΑΓ΋༏Ε͍ͯΔ͜ͱΛ࣮ূ͠· ͨ͠ɽίʔυ͸ɼʜ
16. • ΤʔδΣϯτɿλεΫΛ෼ղͯ͠ɼ໨ඪʹ޲͚ͯߦಈΛબͿɽ • ୅දతͳ 3F"DU ͸ɼࢥߟͱߦಈΛަޓʹߦ͏ɽ[Yao+, ICLR, 2023] ஌తࣗ཯γεςϜͷΤʔδΣϯτ 15
17. • --.ʹɼ࣮૷໨తΛೖྗͨ͠Β࣮૷ίʔυͱ݁Ռ͕ಘΒΕΔɽ • ྑ͍݁ՌΛಘΔʹ͸ɼ޻෉͕ඞཁɽ • --.
    ͕ೖྗͷઐ໳༻ޠΛཧղͰ͖ͳ͍ɽˠ
    3"(
    ʹΑΔ஌֦ࣝॆ • --.
    ͷग़ྗܗ͕ࣜࢦࣔ௨Γͱ͸ҟͳΔɽˠ
    ߏ଄Խग़ྗʹΑΔܗ੍ࣜݶ • ίʔυ͸ಘΒΕ͕ͨɼಈ͔ͳ͍ɽˠ
    --.ʹΑΔπʔϧ࢖༻ • ҙਤ௨Γͷ࣮૷Λͤͣɼྫ֎ॲཧͷΈΛ࣮૷͢Δɽˠ
    ΤʔδΣϯτʹΑΔ֬ೝ • ʜ --.ʹΑΔ৽੍ޚख๏ͷ࣮૷ίʔσΟϯάΛ࣮ݱ͢Δʹ͸ 16 ʻೖྗʼख๏"ΑΓ΋ߴੑೳͳख๏ ΛఏҊ͢Δɽର৅ܥ͸ɼϓϩηε9 ͱ͠ɼγϛϡϨʔγϣϯͰ༗༻ੑ Λࣔ͢ɽ --. ʻग़ྗʼϓϩηε9Λର৅ͱͨ͠ ࣮ݧʹ͓͍ͯɼఏҊख๏͕ख๏" ΑΓ΋༏Ε͍ͯΔ͜ͱΛ࣮ূ͠· ͨ͠ɽίʔυ͸ɼʜ
18. 
    ୈճ৽ϓϩηε੍ޚݚڀձࢿྉΑΓ ϓϩϯϓτΤϯδχΞϦϯά΍ࡐྉ։ൃͳͲͷݚڀ͕͋Δ͕ɼ "*ؔ࿈ͷֶձ΄Ͳ੝Γ্͕͍ͬͯͳ͍ɽ • ೥ͷ "*$I&
    "OOVBM
    .FFUJOH
    ʹ͓͍ͯɼMBSHF
    MBOHVBHF
    NPEFMʢ--.ʣ ΛλΠτϧʹؚΉൃද͸݅ɽҎԼ͸ͦͷҰ෦ɽ • 1SPNQU
    &OHJOFFSJOH%JSFDUFE
    JO$POUFYU
    -FBSOJOH
    PG
    BO
    --. XJUI
    *OUFHSBUFE
    "VUPNBUFE
    4JNVMBUJPO
    GPS
    (FOFSBUJWF
    $IFNJDBM
    1SPDFTT • "
    -BSHF
    -BOHVBHF
    .PEFM
    GPS
    $IFNJDBM
    1SPDFTT
    &OHJOFFSJOH
    1*-05 • *OUFHSBUJOH
    $IFNJTUSZ
    ,OPXMFEHF
    JO
    -BSHF
    -BOHVBHF
    .PEFMT
    7JB
    1SPNQU
    &OHJOFFSJOH • *OUFSQSFUBCMF
    'BVMU
    %FUFDUJPO
    JO
    $IFNJDBM
    1SPDFTTFT
    6TJOH
    -BSHF
    -BOHVBHF
    .PEFMT • -BSHF
    -BOHVBHF
    .PEFMT GPS
    %JTDPWFSJOH
    &RVBUJPOT • -BSHF
    -BOHVBHF
    .PEFM BOE
    .VMUJNPEBM
    -FBSOJOH
    'SBNFXPSL
    GPS
    $BUBMZTU
    %JTDPWFSZ Խֶ޻ֶ෼໺ʹ͓͚Δ--.ͷར༻ 17
19. • --.
    ͕ೖྗͷઐ໳༻ޠΛཧղͰ͖ͳ͍ɽˠ
    3"(
    ʹΑΔ஌֦ࣝॆ • Φϯτϩδʔʢ0OUP$"1& [Morbach+, Eng. Appl. Artif. Intell., 2007]ʣͷ׆༻ • 14&
    ෼໺ಛԽܕݴޠϞσϧ [Kato+, IFAC-PapersOnLine, 2022] ͷར༻ • --.
    ͷग़ྗܗ͕ࣜࢦࣔ௨Γͱ͸ҟͳΔɽˠ
    ߏ଄Խग़ྗʹΑΔܗ੍ࣜݶ • Խֶ޻ֶత஌ࣝʹج੍ͮ͘໿ɿ ୯Ґɼ࣍ݩɼऩࢧɼਤͱͷ੔߹ੑʜ • ίʔυ͸ಘΒΕ͕ͨɼಈ͔ͳ͍ɽˠ
    --.ʹΑΔπʔϧ࢖༻ • 14&
    ಛԽͷ࣮ߦ؀ڥʢྫ͑͹ɼϓϩηεγϛϡϨʔλʣͱͷ઀ଓ • ҙਤ௨Γͷ࣮૷Λͤͣɼྫ֎ॲཧͷΈΛ࣮૷͢Δɽˠ
    ΤʔδΣϯτʹΑΔ֬ೝ • ্هͷٕज़Λू໿ͨ͠ϚϧνΤʔδΣϯτγεςϜ 14&෼໺Ͱ--.Λ࢖͏ࡍͷ໰୊఺ͱղܾࡦ 18
20. • 14&ʹ͓͚Δ--.׆༻ͷϘτϧωοΫΛ੔ཧ͠ɼ ʮ࣭໰Ԡ౴ʯͰ͸ͳ͘࠶ར༻ՄೳͳҰൠϞδϡʔϧੜ੒Λ໨ࢦ͢࿮૊ΈΛఏҊɽ • *OTUSVDUJPO
    
    $POUFYU
    
    *OQVU
    
    0VUQVU
    ʹΑΔඪ४Խϓϩϯϓτߏ଄ɼ UBTL
    DIBJOJOHʹΑΔஈ֊తλεΫ෼ղɼ JUFSBUJWF
    HFOFSBUJPOʹΑΔର࿩తվળͰɼग़ྗͷ҆ఆੑͱଥ౰ੑΛ޲্ɽ • γεςϜ෼ղɼϑϩʔγʔτॱং෇͚ɼϑϩʔγʔτՄࢹԽɼ ϓϩηεγϛϡϨʔγϣϯͳͲͷ14&ؔ࿈λεΫͰ࣮ূɽ 14&෼໺Ͱͷ--.׆༻ྫ̍ 19 [Tao+, Comput. Chem. Eng., 2025]
21. • Խֶϓϩηε҆શͷ஌ࣝάϥϑʴ (SBQI
    3"(
    Ͱ
    --.
    Λ֦ுɽ Խֶ޻ۀஂ஍ࣄނͷۓٸରԠखॱΛ ࣗಈੜ੒ɾධՁͨ͠ɽ • $IFNJDBM
    QSPDFTT
    TBGFUZ
    LOPXMFEHF
    HSBQI
    $14,( ʴ(SBQI
    3"(ʴ--.
    • ର৅෼໺ͷΦϯτϩδʹج͖ͮɼ ؂ࢹσʔλɾࣄނใࠂɾ๏نɾܭըॻ ͔Β
    $14,(
    Λߏங͠ɼ(SBQI
    3"(
    Ͱؔ࿈αϒάϥϑΛऔಘͯ͠--.
    ʹ ༩͑ɺۓٸରԠखॱΛੜ੒͢Δɽ 14&෼໺Ͱͷ--.׆༻ྫ̎ 20 [Zheng+, Can. J. Chem. Eng., 2025]
22. • ϓϩηε੍ޚͱࣗવݴޠॲཧΛ૊Έ߹Θͤͨݚڀ͸ݶఆతͳ͕Βଘࡏ • ओʹʮӡసࢧԉʯʮҟৗରԠʯʮઃܭࢧԉʯʮ҆શࢧԉʯͷΧςΰϦʹ෼ྨՄೳ • ٕज़ͷओ࣠͸ɺ෼ྨɾநग़ɾཁ໿ɾੜ੒ͱ͍ͬͨ/-1ͷجຊٕज़ͱͦͷ--.ల։ ϓϩηε੍ޚ Y
    ࣗવݴޠॲཧͷطଘݚڀ 21 ӡసࢧԉ ҟৗରԠ ੍ޚઃܭࢧԉ ҆શղੳ ϓϩηε੍ޚ º ࣗવݴޠॲཧʢ/-1جຊٕज़ͱ--.ʣ ෼ྨ நग़ ཁ໿ ੜ੒
23. • 3VMFCBTFEͷεϖΠϯޠૢ࡞໋ྩΛ੍ޚ৴߸ʹม׵ [Romano, IFAC Proc. Vol., 1992] • --.ʢ$IBU(-.
    --B."ʣͱ#&35Λ૊Έ߹Θͤͨ࢈ۀ2"γεςϜ <-JV
    &OHJOFFSJOH
    > • "7&7"
    1SPDFTT
    4JNVMBUJPOͱ--.ΤʔδΣϯτΛ.$1Ͱͭͳ͗ɼର࿩తɾ ൒ࣗ཯తʹϓϩηεγϛϡϨʔγϣϯΛߦ͏࿮૊ΈͷఏҊ <-JBOH
    BS9JW
    > ӡసࢧԉɿࣗવݴޠʹΑΔૢ࡞ɾ࣭໰Ԡ౴ 22 [Romano+, IFAC Proc. Vol., 1992] [Liang+, arXiv, 2026]
24. • ϓϥϯτͷ๲େͳΞϥʔϜΛର৅ʹɼϚϧνϥϕϧނো෼ྨϑϨʔϜϫʔΫΛߏங [Shatewakasi+, IECON, 2024] • อશϩά͔Βͷ਍அࢧԉʹ--.Λར༻ [Jose+, Expert Syst. Appl., 2024] ҟৗରԠɿΞϥʔϜղੳͱ਍அࢧԉ 23 [Jose+, Expert Syst. Appl., 2024] [Shatewakasi+, IECON, 2024]
25. • 1'%ΛςΩετԽ͠ɺ੍ޚߏ଄Λ--.Ͱࣗಈੜ੒ [Hirtreiter+, AIChE J. 2024] • ཁ݅࢓༷͔Βߏ੒ཁૉநग़͠ɼ1-$
    1SPHSBNNBCMF
    -PHJD
    $POUSPMMFS
    1-$ ίʔυΛੜ੒ [Boudribila+, Ing. Syst. D. Inf., 2025] ઃܭࢧԉɿ1'%ˠ੍ޚߏ଄ɼςΩετˠ1-$ 24 [Hirtreiter+, AIChE J. 2024]
26. • )";01จॻͷ෼ྨɾ৘ใநग़ [Feng+, Process Saf. Environ. Prot., 2021] • ஌ࣝάϥϑΛ--.ͷࠜڌͱͯ͠׆༻ [Zheng+, Can. J. Chem. Eng., 2025] ҆શɾ஌ࣝ؅ཧɿ)";01ղੳͱจॻφϨοδͷ׆༻ 25 [Zheng+, Can. J. Chem. Eng., 2025] [Feng+, Process Saf. Environ. Prot., 2021]
27. • Ԡ༻ྖҬ͸ͭɺ/-1ٕज़͸෼ྨɾநग़ɾੜ੒ɾର࿩΁޿͕͍ͬͯΔ • ݱঢ়ͷ੍໿͸σʔλʢٕज़จॻɾϩάʣͷ੔උͱ࣮૷ͷݕূίετ • ࠓޙ͸ɺ--.ʴݕূ૚ʢ੍໿ɾ)*5-ʣʹΑΔࢧԉ͕ظ଴͞ΕΔʁ ૯ׅɿ1SPDFTT
    $POUSPM
    º /-1ͷల๬ͱ՝୊ 26 ӡసࢧԉ ҟৗରԠ ੍ޚઃܭࢧԉ ҆શղੳ ϓϩηε੍ޚ º ࣗવݴޠॲཧʢ/-1جຊٕज़ͱ--.ʣ º "*ٕज़ ෼ྨ நग़ ཁ໿ ੜ੒ σʔλ੔උ ʢ࣮σʔλɾΦϯτϩδʣ υϝΠϯಛԽܕLLM ڧԽֶश ࿈߹ֶश ʢFederated learningʣ
28. [Zhao+, arXiv, 2024] W. X. Zhao et al., “A Survey of Large Language Models.” arXiv, 2024. [Liu+, ACM Comput. Surv., 2023]: P. Liu et al., “Pre-train, Prompt, and Predict: A Systematic Survey of Prompting Methods in Natural Language Processing.” ACM Comput. Surv. 55, 9, 2023. [Brown+, NeurIPS, 2020] T. B. Brown et al., “Language Models are Few-Shot Learners.” NeurIPS, 1877–1901, 2020. [Wei+, NeurIPS, 2022] J. Wei et al., “Chain-of-Thought Prompting Elicits Reasoning in Large Language Models.” NeurIPS, 24824–24837, 2022. [Wang+, ICLR, 2023] X. Wang et al., “Self-Consistency Improves Chain of Thought Reasoning in Language Models.” ICLR, 2023. [Lu+, arXiv, 2024] C. Lu et al., “The AI ScientistɿTowards Fully Automated Open-Ended Scientific Discovery.” arXiv, 2024. [Lewis+, NeurIPS, 2020] P. Lewis et al. “Retrieval-augmented generation for knowledge-intensive NLP tasks.” NeurIPS, 2020. [Geng+, arXiv, 2025] S. Geng et al. “JSONSchemaBench: A rigorous benchmark of structured outputs for language models." arXiv, 2025. [Guo+, ACL, 2024] Z. Guo et al., "StableToolBench: Towards Stable Large-Scale Benchmarking on Tool Learning of Large Language Model." ACL, 2024. [Yao+, ICLR, 2023] S. Yao et al., "ReAct: Synergizing Reasoning and Acting in Language Models." ICLR, 2023. ࢀߟจݙ 27
29. [Morbach+, Eng. Appl. Artif. Intell., 2007] J. Morbach et al., "OntoCAPE-A large-scale ontology for chemical process engineering." Eng. Appl. Artif. Intell., 20, 2, 2007. [Kato+, IFAC-PapersOnLine, 2022] S. Kato et al., "ProcessBERT: A Pre-trained Language Model for Judging Equivalence of Variable Definitions in Process Models." IFAC-PapersOnLine, 55, 7, 2022. [Tao+, Comput. Chem. Eng., 2025] X. Tao et al., "From prompt design to iterative generation: Leveraging LLMs in PSE applications." Comput. Chem. Eng., 202, 2025. [Zheng+, Can. J. Che. Eng., 2025] C. Zheng et al., “Chemical process safety domain knowledge graph-enhanced LLM for efficient emergency response decision support.” Can. J. Che. Eng., 103, 10, 2025. [Romano+, IFAC Proc. Vol., 1992] J.M.G. Romano et al., “Natural language interface in control application to a pilot plant.” IFAC Proc. Vol, 25, 6, 1992. [Liu+, Engineering, 2025] R. Liu et al. “Knowledge enhanced industrial question-answering using large language models.” Engineering (Beijing, China), 2025. [Liang+, arXiv, 2026] J. Liang et al., “Large language model agent for user-friendly chemical process simulations.” arXiv, 2025. [Shatewakasi+, IECON, 2024] Y. Shatewakasi et al., “Towards NLP-driven online-classification of industrial alarm messages.” IECON, 2024. [Jose+, Expert Syst. Appl., 2024] S. Jose et al., “Advancing multimodal diagnostics: Integrating industrial textual data and domain knowledge with large language models.” Expert Syst. Appl., 255, 2024. ࢀߟจݙ 28
30. [Hirtreiter+, AIChE J. 2024] E. Hirtreiter et al., “Toward automatic generation of control structures for process flow diagrams with large language models.” AIChE J. 70, 1, 2024. [Boudribila+, Ing. Syst. D. Inf., 2025] A. Boudribila et al., “Automatic generation of PLC control code from natural language requirement specifications.” Ing. Syst. D. Inf., 30, 6, 2025. [Feng+, Process Saf. Environ. Prot., 2021] X. Feng et al., “Application of natural language processing in HAZOP reports.” Process Saf. Environ. Prot., 155, 2021. ࢀߟจݙ 29