demo / multi /index.html
Hodfa71's picture
Multi-turn: add v2 combined (all turns in one clip)
00e2586 verified
Raw
History Blame Contribute Delete
4.88 kB
<!doctype html>
<html lang="en"><head><meta charset="utf-8">
<meta name="viewport" content="width=device-width, initial-scale=1">
<title>Multi-turn: both deliveries (actual released audio)</title>
<style>
:root { color-scheme: light dark; }
body { font-family: system-ui,-apple-system,Segoe UI,Roboto,sans-serif; max-width: 860px;
margin: 2rem auto; padding: 0 1rem; line-height: 1.5; }
h1 { font-size: 1.4rem; } .lead { opacity: .82; }
h2 { margin: 1.8rem 0 .4rem; font-size: 1.02rem; }
.v1 { background: rgba(127,127,127,.12); padding: .6rem .9rem; border-radius: 8px; margin: .4rem 0 .8rem; }
.turns { display: flex; flex-wrap: wrap; gap: .5rem; }
.turn { display: flex; flex-direction: column; gap: .2rem; }
.tl { font-size: .75rem; font-weight: 600; opacity: .8; }
audio { width: 180px; height: 32px; }
.tag { font-size: .72rem; opacity: .65; }
</style></head><body>
<h1>Multi-turn: both deliveries, actual released audio</h1>
<p class="lead">Straight from the dataset (omniagentbench/OmniAgentBench/wild_long_scattered/mpcc/audio_tts_qwen3/2B). <strong>v1</strong> is the
single concatenated clip the evaluation used. <strong>v2</strong> is the same
instruction as separate per-turn clips, which the fixed engine can feed as a real
conversation. Both files already ship in the release.</p>
<div id="root"></div>
<script>
const M = {"source": "omniagentbench/OmniAgentBench/wild_long_scattered/mpcc/audio_tts_qwen3/2B", "samples": [{"sample_id": "mpcc_flight_medium_1", "n_turns": 7, "turns": [{"idx": 1, "audio": "actual/mpcc_flight_medium_1/turn1.wav"}, {"idx": 2, "audio": "actual/mpcc_flight_medium_1/turn2.wav"}, {"idx": 3, "audio": "actual/mpcc_flight_medium_1/turn3.wav"}, {"idx": 4, "audio": "actual/mpcc_flight_medium_1/turn4.wav"}, {"idx": 5, "audio": "actual/mpcc_flight_medium_1/turn5.wav"}, {"idx": 6, "audio": "actual/mpcc_flight_medium_1/turn6.wav"}, {"idx": 7, "audio": "actual/mpcc_flight_medium_1/turn7.wav"}], "concat": "actual/mpcc_flight_medium_1/concat.wav", "v2_combined": "actual/mpcc_flight_medium_1/v2_combined.wav"}, {"sample_id": "mpcc_calendar_medium_1", "n_turns": 7, "turns": [{"idx": 1, "audio": "actual/mpcc_calendar_medium_1/turn1.wav"}, {"idx": 2, "audio": "actual/mpcc_calendar_medium_1/turn2.wav"}, {"idx": 3, "audio": "actual/mpcc_calendar_medium_1/turn3.wav"}, {"idx": 4, "audio": "actual/mpcc_calendar_medium_1/turn4.wav"}, {"idx": 5, "audio": "actual/mpcc_calendar_medium_1/turn5.wav"}, {"idx": 6, "audio": "actual/mpcc_calendar_medium_1/turn6.wav"}, {"idx": 7, "audio": "actual/mpcc_calendar_medium_1/turn7.wav"}], "concat": "actual/mpcc_calendar_medium_1/concat.wav", "v2_combined": "actual/mpcc_calendar_medium_1/v2_combined.wav"}, {"sample_id": "mpcc_meeting_medium_1", "n_turns": 7, "turns": [{"idx": 1, "audio": "actual/mpcc_meeting_medium_1/turn1.wav"}, {"idx": 2, "audio": "actual/mpcc_meeting_medium_1/turn2.wav"}, {"idx": 3, "audio": "actual/mpcc_meeting_medium_1/turn3.wav"}, {"idx": 4, "audio": "actual/mpcc_meeting_medium_1/turn4.wav"}, {"idx": 5, "audio": "actual/mpcc_meeting_medium_1/turn5.wav"}, {"idx": 6, "audio": "actual/mpcc_meeting_medium_1/turn6.wav"}, {"idx": 7, "audio": "actual/mpcc_meeting_medium_1/turn7.wav"}], "concat": "actual/mpcc_meeting_medium_1/concat.wav", "v2_combined": "actual/mpcc_meeting_medium_1/v2_combined.wav"}]};
const root = document.getElementById("root");
for (const s of M.samples) {
const h = document.createElement("h2");
h.textContent = s.sample_id + " (" + s.n_turns + " turns)";
root.appendChild(h);
if (s.concat) {
const d = document.createElement("div"); d.className = "v1";
d.innerHTML = '<span class="tag">v1 concatenated (one clip)</span><br>';
const a = document.createElement("audio"); a.controls = true; a.preload = "none"; a.src = s.concat;
d.appendChild(a); root.appendChild(d);
}
if (s.v2_combined) {
const d = document.createElement("div"); d.className = "v1";
d.innerHTML = '<span class="tag">v2 combined (all turns, one clip, 0.5 s gaps)</span><br>';
const a = document.createElement("audio"); a.controls = true; a.preload = "none"; a.src = s.v2_combined;
d.appendChild(a); root.appendChild(d);
}
const wrap = document.createElement("div");
const lbl = document.createElement("div"); lbl.className = "tag"; lbl.textContent = "v2 per-turn (separate clips)";
wrap.appendChild(lbl);
const row = document.createElement("div"); row.className = "turns";
for (const t of s.turns) {
const c = document.createElement("div"); c.className = "turn";
const l = document.createElement("div"); l.className = "tl"; l.textContent = "T" + t.idx; c.appendChild(l);
const a = document.createElement("audio"); a.controls = true; a.preload = "none"; a.src = t.audio; c.appendChild(a);
row.appendChild(c);
}
wrap.appendChild(row); root.appendChild(wrap);
}
</script></body></html>