As an orchestration infrastructure, agentic AI systems now can plan and decompose the pre-defined goals into a sequence of steps, decide on external function calls, and coordinate one LLM or multiple LLMs with specialised roles. In this context, this position paper adopts a future studies “backcasting” ap-proach that starts with a desirable future, en-visioned as one in which AI-integrated trans-lation workflows are transparent, accountable, and aligned with human values; it then works backwards to examine how translator exper-tise should be reconceptualised to sustain meaningful human-in-the-loop participation. In this sense, the study first conceptualises the current translation-service provision as a sys-tem structured around managerial, mediation, and authorising roles. It then analyses how these roles may be changed and augmented within the agentic AI-orchestrated workflows. Building on the analysis, we propose a series of competences that should be cultivated to achieve the envisioned future: 1) evaluation grounded in advanced language competence; 2) situated and context-sensitive judgement informed by cultural and experiential knowledge; and 3) strategic procedural plan-ning in the design and oversight of agentic AI-orchestration workflows. The paper con-cludes with recommendations for future peda-gogical development and empirical research.

Recent work on non-autoregressive neural machine translation (NAT) that leverages alignment information to explicitly reduce the modality of target distribution has reported comparable performance with counterparts that tackle multi-modality problem by implicitly modeling dependencies. Effectiveness in handling alignment is vital for models that follow this approach, where a token reordering mechanism is typically involved and plays a vital role. We review the reordering capability of the respective mechanisms in recent NAT models, and our experimental results show that their performance is sub-optimal. We propose to learn a non-autoregressive language model (NALM) based on transformer which can be combined with Viterbi decoding to achieve better reordering performance. We evaluate the proposed NALM using the PTB dataset where sentences with words permuted in different ways are expected to have their ordering recovered. Our empirical results show that the proposed method can outperform the state-of-the-art reordering mechanisms under different word permutation settings, with a 2-27 BLEU improvement, suggesting high potential for word alignment in NAT.