{{ ui_lang.arena_lede|default('Send one prompt to every configured model side by side, compare the responses, and pick the best — building a Bradley-Terry & Elo leaderboard and exportable DPO preference data. Models are provider-agnostic (any of Potato\'s LLM endpoints).') }}

{{ ui_lang.arena_sent_to|default('Sent to') }} {{ models|length }} {{ ui_lang.arena_models_label|default('model(s):') }} {{ models|join(', ') }}

{{ ui_lang.arena_lb_note|default('Ranked by Bradley-Terry score (accounts for opponent strength), with Elo ratings. Win rate shown for reference.')|safe }}

{% for r in leaderboard %} {% endfor %}

{{ ui_lang.arena_th_model\|default('Model') }}	{{ ui_lang.arena_th_bt\|default('BT score') }}	{{ ui_lang.arena_th_elo\|default('Elo') }}	{{ ui_lang.arena_th_wins\|default('Wins') }}	{{ ui_lang.arena_th_comparisons\|default('Comparisons') }}	{{ ui_lang.arena_th_win_rate\|default('Win rate') }}
{{ r.label }}	{{ '%.1f'\|format(r.bt_score) if r.bt_score is not none else '—' }}	{{ r.elo if r.elo is not none else '—' }}	{{ r.wins }}	{{ r.comparisons }}	{{ '%.0f%%'\|format(r.win_rate100) if r.win_rate is not none else '—' }}{% if r.win_rate_lo is not none %}{{ '%.0f–%.0f%%'\|format(r.win_rate_lo100, r.win_rate_hi*100) }}{% endif %}

{{ ui_lang.arena_heading|default('Model Arena') }}

{{ ui_lang.arena_prompt_heading|default('Prompt') }}

{{ ui_lang.arena_responses_heading|default('Responses') }}

{{ ui_lang.arena_leaderboard_heading|default('Leaderboard') }}