Mechanism

Transfers behaviors that initially require a long in-context system prompt, persona specification, constitutional rulebook, or few-shot scratchpad cc into a zero-prompt student πθ(y∣x)\pi_\theta(y\mid x) by minimizing token-level KL divergence DKL(πteacher(y∣c,x) ∥ πθ(y∣x))\mathbb{D}_{\text{KL}}(\pi_{\text{teacher}}(y\mid c, x) \,\|\, \pi_\theta(y\mid x)) (or slicing out the conditioning prefix cc after turn 1 as in Llama 2's Ghost Attention (GAtt) so multi-turn attention stays anchored to the system prompt).

References

  1. A General Language Assistant as a Laboratory for Alignment ↗

    Amanda Askell, Yuntao Bai, Anna Chen, Dawn Drain, Deep Ganguli, Tom Henighan et al. · 2021

    All 22 authors

    Amanda Askell, Yuntao Bai, Anna Chen, Dawn Drain, Deep Ganguli, Tom Henighan, Andy Jones, Nicholas Joseph, Ben Mann, Nova DasSarma, Nelson Elhage, Zac Hatfield-Dodds, Danny Hernandez, Jackson Kernion, Kamal Ndousse, Catherine Olsson, Dario Amodei, Tom Brown, Jack Clark, Sam McCandlish, Chris Olah, Jared Kaplan

  2. Learning by Distilling Context ↗

    Charlie Snell, Dan Klein, Ruiqi Zhong · 2022

  3. Llama 2: Open Foundation and Fine-Tuned Chat Models ↗

    Touvron, Hugo, Martin, Louis, Stone, Kevin, Albert, Peter, Almahairi, Amjad, Babaei, Yasmine et al. · 2023

    All 68 authors

    Touvron, Hugo, Martin, Louis, Stone, Kevin, Albert, Peter, Almahairi, Amjad, Babaei, Yasmine, Bashlykov, Nikolay, Batra, Soumya, Bhargava, Prajjwal, Bhosale, Shruti, Bikel, Dan, Blecher, Lukas, Ferrer, Cristian Canton, Chen, Moya, Cucurull, Guillem, Esiobu, David, Fernandes, Jude, Fu, Jeremy, Fu, Wenyin, Fuller, Brian, Gao, Cynthia, Goswami, Vedanuj, Goyal, Naman, Hartshorn, Anthony, Hosseini, Saghar, Hou, Rui, Inan, Hakan, Kardas, Marcin, Kerkez, Viktor, Khabsa, Madian, Kloumann, Isabel, Korenev, Artem, Koura, Punit Singh, Lachaux, Marie-Anne, Lavril, Thibaut, Lee, Jenya, Liskovich, Diana, Lu, Yinghai, Mao, Yuning, Martinet, Xavier, Mihaylov, Todor, Mishra, Pushkar, Molybog, Igor, Nie, Yixin, Poulton, Andrew, Reizenstein, Jeremy, Rungta, Rashi, Saladi, Kalyan, Schelten, Alan, Silva, Ruan, Smith, Eric Michael, Subramanian, Ranjan, Tan, Xiaoqing Ellen, Tang, Binh, Taylor, Ross, Williams, Adina, Kuan, Jian Xiang, Xu, Puxin, Yan, Zheng, Zarov, Iliyan, Zhang, Yuchen, Fan, Angela, Kambadur, Melanie, Narang, Sharan, Rodriguez, Aurelien, Stojnic, Robert, Edunov, Sergey, Scialom, Thomas