Spaces:
Running
Running
File size: 25,877 Bytes
268dd75 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 | // SPDX-License-Identifier: Apache-2.0
// © 2026 Lutar, Stephen P. — SZL Holdings · ORCID 0009-0001-0110-4173 · Doctrine v11
//
// surfaces/nested.js — NESTED LEARNING / CONTINUUM MEMORY SYSTEM ("Hope")
// organ for the holographic frontier ring (Behrouz, Razaviyayn, Zhong &
// Mirrokni, Google Research, NeurIPS 2025 — "the illusion of deep learning
// architectures"). Renders a THREE-TIER PLASTICITY LADDER of memory levels at
// different update clocks (fast k=1, medium k=8, slow k=64), fed by a streaming
// toy continual-learning task-block sequence. A HUD contrasts the FORGETTING
// CURVE of task A for a single-clock baseline vs the multi-timescale CMS
// schedule from the live snapshot at /api/killinchu/v1/nested/schedule. The
// slow "core" (protected by a surprise gate) glows proof-teal as it retains
// early tasks; the always-on baseline overwrites and forgets (grey). Honesty
// label "MODELED" is read VERBATIM from the JSON and displayed as-is; never
// upgraded.
//
// Surface export shape (mirrors titans.js / episodic.js exactly):
// export default { id, title, endpoints, mount(ctx), unmount() }
// ctx = { stage, container, live, label, THREE, szl3d }
//
// DATA SHOWN (all from live endpoint, inside payload):
// num_tasks, block_len, n_tokens, k_fast, k_med, k_slow, gate, levels[],
// taskA_baseline[], taskA_cms[], retainedA_baseline, retainedA_cms,
// retention_delta, mean_retention_delta, slow_writes_cms,
// slow_writes_baseline, mean_surprise, peak_surprise, momentum, forget_curve[]
//
// LEADERS ADOPTED & CITED (clean-room; NOT claimed as SZL's own):
// Nested Learning: The Illusion of Deep Learning Architectures (the paradigm,
// Continuum Memory System, and "Hope" proof-of-concept simulated here):
// Behrouz, Razaviyayn, Zhong & Mirrokni 2025, Google Research, NeurIPS 2025
// https://arxiv.org/abs/2512.24695
// Google Research blog — Introducing Nested Learning (official announcement):
// https://research.google/blog/introducing-nested-learning-a-new-ml-paradigm-for-continual-learning/
//
// HONESTY LABELS: MODELED (deterministic re-implementation of the multi-
// timescale update-scheduling + optimizer-as-memory arithmetic; NOT the Hope
// architecture; NEVER-CLAIMED-AS a trained model). Independent-ablation
// nuance: the NeurIPS paper was noted in public commentary to lack full
// ablations in its first version, so this demonstrates the SCHEDULING
// MECHANISM reduces forgetting on a controlled toy stream — NOT reproducing
// Hope's benchmark wins. Read verbatim from JSON; never upgraded here. The
// endpoint nests its fields under `payload`, with the label at the top level —
// this surface handles the label at top-level OR inside payload.label
// defensively.
// COLOURS: lattice-blue 0x5b8dee (fast level / task stream / spine), violet-blue
// 0x8a6bff (medium level / clock ring), proof-teal 0x3af4c8 (protected slow
// core / retention / HUD accent), greys (forgotten / degraded state). Purple
// BANNED.
// 0 RUNTIME CDN. three.js via ctx.THREE (vendored by the page importmap).
// DOCTRINE v11: degrades gracefully (grey) on 404/error; honesty label still shown.
// Nothing here is in the locked-8. Λ stays Conjecture 1. Trust never 100%.
const ID = "nested";
const TITLE = "Nested Learning · Continuum Memory System (live)";
// Endpoint is hosted on the dedicated killinchu Space (isolated compute), reached
// cross-origin (killinchu returns access-control-allow-origin: https://a-11-oy.com).
// This keeps the nested organ's rebuilds/faults isolated from the flagship.
const EP = "https://szlholdings-killinchu.hf.space/api/killinchu/v1/nested/schedule?seed=42&num_tasks=6&k_fast=1&k_med=8&k_slow=64&gate=0.15";
// data-viz hues — purple BANNED
const C_FAST = 0x5b8dee; // lattice-blue (fast level / task stream / spine)
const C_MED = 0x8a6bff; // violet-blue (medium level / clock ring)
const C_SLOW = 0x3af4c8; // proof-teal (protected slow core / retention / HUD accent)
const C_FORGET = 0x5a6570; // grey (forgotten / overwritten baseline point)
const C_DIM = 0x42505d; // grey (degraded / no-live-data)
const C_GRID = 0x1b3a44; // floor / link colour
// layout geometry
const LEVEL_Y = [0.7, 2.2, 3.7]; // y of fast / med / slow tiers (bottom -> top)
const LEVEL_R = [1.4, 2.6, 3.8]; // ring radius of each tier
const RING_SEG = 64; // ring outline segments
const MAX_BLOCKS = 96; // cap on forgetting-curve markers rendered
const CURVE_SPAN = 12.0; // world-units the forgetting curve spans along X
let _stage = null, _THREE = null, _ctx = null, _group = null, _overlay = null;
let _frameReg = false, _polls = [], _el = {}, _badge = null;
let _plain = false;
// geometry handles
let _floor = null;
let _rings = []; // Array<THREE.LineLoop> — one clock ring per level
let _tiers = []; // Array<THREE.Mesh> — the three memory-level nodes
let _baseMarks = []; // Array<THREE.Mesh> — baseline forgetting-curve markers
let _cmsMarks = []; // Array<THREE.Mesh> — CMS forgetting-curve markers
let _core = null; // THREE.Mesh — central slow-core "retention" node
// live state
const S = {
label: null,
numTasks: null, // num_tasks
blockLen: null, // block_len
nTokens: null, // n_tokens
kFast: null, // k_fast
kMed: null, // k_med
kSlow: null, // k_slow
gate: null, // gate
taskABase: null, // taskA_baseline[]
taskACms: null, // taskA_cms[]
retainBase: null, // retainedA_baseline
retainCms: null, // retainedA_cms
retDelta: null, // retention_delta
meanDelta: null, // mean_retention_delta
slowWritesCms: null, // slow_writes_cms
slowWritesBase: null, // slow_writes_baseline
meanSurprise: null, // mean_surprise
peakSurprise: null, // peak_surprise
momentum: null, // momentum
curve: null, // forget_curve[]
state: "init",
};
// =============================================================================
// mount(ctx)
// =============================================================================
export function mount(ctx) {
_ctx = ctx; _stage = ctx.stage; _THREE = ctx.THREE;
_group = new _THREE.Group();
_stage.scene.add(_group);
_stage.camera.position.set(0, 8, 20);
try { if (_stage.controls && _stage.controls.target) { _stage.controls.target.set(0, 2.0, 0); _stage.controls.update(); } } catch (_) {}
try { _stage.setBloom(true); } catch (_) {}
_buildFloor();
_buildLadder();
_buildCurve();
_buildCore();
if (!_frameReg) { _stage.onFrame(_onFrame); _frameReg = true; }
_badge = ctx.live.createBadge();
_polls.push(ctx.live.poll(EP, 5000, _onNested, { badge: _badge, onState: (m) => { S.state = m.state; _paintOverlay(); } }));
_buildOverlay();
return { id: ID, started: true };
}
// =============================================================================
// builders
// =============================================================================
function _buildFloor() {
const THREE = _THREE;
const grid = new THREE.GridHelper(44, 44, C_GRID, 0x0f2027);
grid.material.opacity = 0.18; grid.material.transparent = true; grid.position.y = -0.01;
_group.add(grid);
_floor = grid;
}
// The plasticity LADDER: three stacked clock rings (fast/med/slow) each with a
// central level node. Higher rings = slower clocks = more protected memory.
function _buildLadder() {
const THREE = _THREE;
const cols = [C_FAST, C_MED, C_SLOW];
for (let lvl = 0; lvl < 3; lvl++) {
const r = LEVEL_R[lvl];
const y = LEVEL_Y[lvl];
// clock ring outline
const pts = [];
for (let i = 0; i <= RING_SEG; i++) {
const a = (i / RING_SEG) * Math.PI * 2;
pts.push(new THREE.Vector3(Math.cos(a) * r, y, Math.sin(a) * r));
}
const geo = new THREE.BufferGeometry().setFromPoints(pts);
const mat = new THREE.LineBasicMaterial({ color: cols[lvl], transparent: true, opacity: 0.4 });
const ring = new THREE.LineLoop(geo, mat);
_group.add(ring);
_rings.push(ring);
// level node (the memory level itself)
const node = new THREE.Mesh(
new THREE.OctahedronGeometry(0.34, 0),
new THREE.MeshStandardMaterial({ color: cols[lvl], emissive: cols[lvl], emissiveIntensity: 0.3, transparent: true, opacity: 0.85 }),
);
node.position.set(0, y, 0);
_group.add(node);
_tiers.push(node);
}
}
// Two rows of forgetting-curve markers streaming along X: task-A accuracy after
// each task block for the single-clock baseline (front row) and the multi-
// timescale CMS schedule (back row). Height encodes retained accuracy.
function _buildCurve() {
const THREE = _THREE;
const mGeo = new THREE.SphereGeometry(0.13, 8, 8);
for (let i = 0; i < MAX_BLOCKS; i++) {
const x = -CURVE_SPAN / 2 + (i / (MAX_BLOCKS - 1)) * CURVE_SPAN;
const b = new THREE.Mesh(
mGeo,
new THREE.MeshStandardMaterial({ color: C_FORGET, emissive: C_FORGET, emissiveIntensity: 0.2, transparent: true, opacity: 0.0 }),
);
b.position.set(x, 0.2, 2.4);
b.visible = false;
_group.add(b);
_baseMarks.push(b);
const c = new THREE.Mesh(
mGeo,
new THREE.MeshStandardMaterial({ color: C_SLOW, emissive: C_SLOW, emissiveIntensity: 0.2, transparent: true, opacity: 0.0 }),
);
c.position.set(x, 0.2, -2.4);
c.visible = false;
_group.add(c);
_cmsMarks.push(c);
}
}
function _buildCore() {
const THREE = _THREE;
_core = new THREE.Mesh(
new THREE.IcosahedronGeometry(0.6, 1),
new THREE.MeshStandardMaterial({ color: C_SLOW, emissive: C_SLOW, emissiveIntensity: 0.45, wireframe: true, transparent: true, opacity: 0.85 }),
);
_core.position.set(0, LEVEL_Y[2], 0);
_group.add(_core);
}
// =============================================================================
// live data handler
// =============================================================================
function _onNested(j) {
// The endpoint nests its metrics under `payload`; the honesty label may sit at
// the TOP LEVEL or (defensively) inside payload.label. Read it VERBATIM from
// wherever it is — never upgrade.
const p = (j && typeof j.payload === "object" && j.payload) ? j.payload : j;
const rawLabel = (j && j.label) || (p && p.label) || "MODELED";
S.label = String(rawLabel).toUpperCase();
S.numTasks = typeof p.num_tasks === "number" ? p.num_tasks : null;
S.blockLen = typeof p.block_len === "number" ? p.block_len : null;
S.nTokens = typeof p.n_tokens === "number" ? p.n_tokens : null;
S.kFast = typeof p.k_fast === "number" ? p.k_fast : null;
S.kMed = typeof p.k_med === "number" ? p.k_med : null;
S.kSlow = typeof p.k_slow === "number" ? p.k_slow : null;
S.gate = typeof p.gate === "number" ? p.gate : null;
S.retainBase = typeof p.retainedA_baseline === "number" ? p.retainedA_baseline : null;
S.retainCms = typeof p.retainedA_cms === "number" ? p.retainedA_cms : null;
S.retDelta = typeof p.retention_delta === "number" ? p.retention_delta : null;
S.meanDelta = typeof p.mean_retention_delta === "number" ? p.mean_retention_delta : null;
S.slowWritesCms = typeof p.slow_writes_cms === "number" ? p.slow_writes_cms : null;
S.slowWritesBase = typeof p.slow_writes_baseline === "number" ? p.slow_writes_baseline : null;
S.meanSurprise = typeof p.mean_surprise === "number" ? p.mean_surprise : null;
S.peakSurprise = typeof p.peak_surprise === "number" ? p.peak_surprise : null;
S.momentum = typeof p.momentum === "number" ? p.momentum : null;
S.taskABase = Array.isArray(p.taskA_baseline) ? p.taskA_baseline : null;
S.taskACms = Array.isArray(p.taskA_cms) ? p.taskA_cms : null;
S.curve = Array.isArray(p.forget_curve) ? p.forget_curve : null;
_updateScene();
_paintOverlay();
}
// =============================================================================
// geometry updater — drives the ladder + forgetting curve from live data
// =============================================================================
function _updateScene() {
const live = S.state === "live";
const cols = [C_FAST, C_MED, C_SLOW];
// clock rings + level nodes degrade to grey when not live
for (let lvl = 0; lvl < 3; lvl++) {
const ring = _rings[lvl];
if (ring) {
ring.material.color.setHex(live ? cols[lvl] : C_DIM);
ring.material.opacity = live ? 0.4 : 0.12;
}
const node = _tiers[lvl];
if (node) {
const c = live ? cols[lvl] : C_FORGET;
node.material.color.setHex(c);
node.material.emissive.setHex(c);
node.material.emissiveIntensity = live ? 0.35 : 0.08;
node.material.opacity = live ? 0.9 : 0.25;
}
}
// forgetting-curve markers: height = retained task-A accuracy per block.
const base = live && S.taskABase && S.taskABase.length ? S.taskABase.slice(0, MAX_BLOCKS) : [];
const cms = live && S.taskACms && S.taskACms.length ? S.taskACms.slice(0, MAX_BLOCKS) : [];
const n = Math.max(base.length, cms.length);
for (let i = 0; i < MAX_BLOCKS; i++) {
const bm = _baseMarks[i];
const cm = _cmsMarks[i];
if (!live || i >= n) { bm.visible = false; cm.visible = false; continue; }
if (i < base.length) {
const a = base[i];
bm.visible = true;
bm.position.y = 0.2 + a * 3.2;
// baseline that has forgotten (low retention) fades to grey
const c = a >= 0.5 ? C_FAST : C_FORGET;
bm.material.color.setHex(c);
bm.material.emissive.setHex(c);
bm.material.emissiveIntensity = 0.2 + 0.5 * a;
bm.material.opacity = 0.35 + 0.55 * a;
bm.scale.setScalar(0.7 + 0.7 * a);
} else { bm.visible = false; }
if (i < cms.length) {
const a = cms[i];
cm.visible = true;
cm.position.y = 0.2 + a * 3.2;
cm.material.color.setHex(C_SLOW);
cm.material.emissive.setHex(C_SLOW);
cm.material.emissiveIntensity = 0.25 + 0.6 * a;
cm.material.opacity = 0.4 + 0.55 * a;
cm.scale.setScalar(0.7 + 0.7 * a);
} else { cm.visible = false; }
}
// slow core: size/brightness reflect how much of task A the protected core
// retains (retainedA_cms), and pulse brighter with the retention advantage.
if (_core) {
if (live && S.retainCms != null) {
_core.material.color.setHex(C_SLOW);
_core.material.emissive.setHex(C_SLOW);
_core.material.opacity = 0.85;
_core.scale.setScalar(0.6 + S.retainCms * 1.0);
} else {
_core.material.color.setHex(C_DIM);
_core.material.emissive.setHex(C_DIM);
_core.material.opacity = 0.3;
_core.scale.setScalar(0.6);
}
}
}
// =============================================================================
// per-frame animation
// =============================================================================
function _onFrame() {
const t = performance.now();
if (_group) _group.rotation.y = Math.sin(t * 0.00008) * 0.14;
// rings spin at their own clock rate — fast ring spins fastest
if (_rings[0]) _rings[0].rotation.y += 0.010;
if (_rings[1]) _rings[1].rotation.y += 0.0035;
if (_rings[2]) _rings[2].rotation.y += 0.0009;
if (_core) {
_core.rotation.y += 0.02;
_core.rotation.x += 0.009;
const pulse = 1.0 + 0.12 * Math.sin(t * 0.0035);
const base = (S.state === "live" && S.retainCms != null) ? (0.6 + S.retainCms * 1.0) : 0.6;
_core.scale.setScalar(base * pulse);
}
}
// =============================================================================
// overlay
// =============================================================================
function _buildOverlay() {
const ctx = _ctx;
_overlay = document.createElement("div");
Object.assign(_overlay.style, {
position: "absolute", left: "14px", top: "14px", zIndex: "6",
display: "flex", flexDirection: "column", gap: "8px",
maxWidth: "min(94%,460px)",
font: "12px ui-sans-serif,system-ui,Segoe UI,Roboto,Arial",
color: "#eef3f6",
});
const h = document.createElement("div");
h.style.cssText = "font:600 13px ui-sans-serif,system-ui;letter-spacing:.4px";
h.textContent = TITLE;
_overlay.appendChild(h);
const sub = document.createElement("div");
sub.style.cssText = "color:#9fb1bf;font-size:11px;line-height:1.55";
sub.innerHTML =
'A model as a stack of memory levels updating at <b>different clocks</b> (fast/medium/slow), with ' +
'the <b>optimizer itself as associative memory</b>. On a toy continual-learning stream, the slow ' +
'<b>protected core</b> (a surprise-gated timescale) <b>retains earlier tasks</b> where a single-clock ' +
'baseline overwrites and <b>forgets</b> them. ' +
'Honesty label <b>MODELED</b> (deterministic scheduling simulation; NOT the Hope architecture). 0 runtime CDN.';
_overlay.appendChild(sub);
const brow = document.createElement("div");
brow.style.cssText = "display:flex;gap:8px;align-items:center;flex-wrap:wrap";
if (_badge && _badge.el) brow.appendChild(_badge.el);
_overlay.appendChild(brow);
const card = document.createElement("div");
card.style.cssText = "background:#0a1117;border:1px solid #1d2a36;border-radius:9px;padding:9px 10px;display:flex;flex-direction:column;gap:6px";
const chead = document.createElement("div");
chead.style.cssText = "display:flex;align-items:center;gap:8px;flex-wrap:wrap";
const dot = document.createElement("span");
dot.style.cssText = "width:9px;height:9px;border-radius:50%;background:#3af4c8;box-shadow:0 0 7px #3af4c8";
const nm = document.createElement("b");
nm.style.cssText = "font-size:12px;color:#3af4c8;letter-spacing:.3px";
nm.textContent = "nested learning / continuum memory system";
chead.appendChild(dot); chead.appendChild(nm);
card.appendChild(chead);
const grid = document.createElement("div");
grid.style.cssText = "display:grid;grid-template-columns:1fr;gap:4px";
function kpiRow(id, label) {
const r = document.createElement("div");
r.style.cssText = "display:flex;justify-content:space-between;gap:10px;font-size:11px";
const l = document.createElement("span"); l.style.cssText = "color:#9fb1bf"; l.textContent = label;
const v = document.createElement("b");
v.id = id;
v.style.cssText = "font-variant-numeric:tabular-nums;color:#eef3f6;text-align:right;max-width:58%";
v.textContent = "\u2014";
_el[id] = v;
r.appendChild(l); r.appendChild(v); return r;
}
grid.appendChild(kpiRow("ne-tasks", "continual-learning tasks"));
grid.appendChild(kpiRow("ne-clocks", "clocks k_fast / k_med / k_slow"));
grid.appendChild(kpiRow("ne-gate", "slow-core surprise gate"));
grid.appendChild(kpiRow("ne-retbase", "task-A retained \u2014 single-clock"));
grid.appendChild(kpiRow("ne-retcms", "task-A retained \u2014 CMS (MODELED)"));
grid.appendChild(kpiRow("ne-delta", "retention gain (CMS \u2212 baseline)"));
grid.appendChild(kpiRow("ne-writes", "slow-core writes CMS / baseline"));
grid.appendChild(kpiRow("ne-surprise","mean / peak surprise"));
grid.appendChild(kpiRow("ne-momentum","momentum (optimizer-as-memory)"));
grid.appendChild(kpiRow("ne-label", "honesty label"));
card.appendChild(grid);
const fn = document.createElement("div");
fn.style.cssText = "font-size:9.5px;color:#6b7a86;line-height:1.5";
fn.textContent = "Behrouz, Razaviyayn, Zhong & Mirrokni 2025 (Google Research) \u00b7 Nested Learning: The Illusion of Deep Learning Architectures \u00b7 NeurIPS 2025 \u00b7 arXiv:2512.24695 \u00b7 research.google/blog Nested Learning. MODELED \u00b7 not claimed-as.";
card.appendChild(fn);
_overlay.appendChild(card);
const pl = document.createElement("button");
pl.textContent = "\u25d1 what this means";
pl.title = "Toggle plain-language explanation for investors & consumers.";
pl.style.cssText = "font:11px ui-monospace,monospace;padding:5px 11px;border-radius:7px;border:1px solid #3af4c8;background:#08140f;color:#3af4c8;cursor:pointer;width:fit-content";
pl.addEventListener("click", () => {
_plain = !_plain;
pl.style.background = _plain ? "#0f2a20" : "#08140f";
_applyPlain();
});
_overlay.appendChild(pl);
const pd = document.createElement("div");
pd.id = "ne-plain";
pd.style.cssText = "font-size:10.5px;color:#c9d6df;line-height:1.55;border:1px dashed #26333f;border-radius:7px;padding:7px 9px;display:none";
_el["plain"] = pd;
_overlay.appendChild(pd);
(ctx.container || document.body).appendChild(_overlay);
_paintOverlay();
}
function _applyPlain() {
const pd = _el["plain"];
if (!pd) return;
pd.style.display = _plain ? "block" : "none";
if (!_plain) return;
const rBase = S.retainBase != null ? (S.retainBase * 100).toFixed(0) + "%" : "loading\u2026";
const rCms = S.retainCms != null ? (S.retainCms * 100).toFixed(0) + "%" : "loading\u2026";
const dPct = S.retDelta != null ? "+" + (S.retDelta * 100).toFixed(0) + " points" : "loading\u2026";
pd.innerHTML =
"<b>What this means:</b> When an AI learns a new skill it often <b>forgets the old one</b> " +
"(\u201ccatastrophic forgetting\u201d). Nested Learning treats the model as a <b>stack of memories that " +
"update at different speeds</b> \u2014 a fast one that chases whatever it is learning right now, and " +
"slower ones that only commit knowledge once it has <b>settled</b>. The slowest \u201ccore\u201d is " +
"<b>protected</b>: it refuses to overwrite itself during the noisy middle of learning something new. " +
"Here we teach it several small tasks in a row and then check how much of the <b>first</b> task it still " +
"remembers. A plain single-speed model keeps only about <b>" + rBase + "</b> of task A, while the " +
"multi-speed schedule keeps about <b>" + rCms + "</b> \u2014 a gain of roughly <b>" + dPct + "</b> less " +
"forgetting on this toy stream. " +
"<b>Honest caveat:</b> this is a <b>MODELED</b> toy simulation of the scheduling idea (tiny linear " +
"tasks, hand-set clocks), <b>not the \u201cHope\u201d model</b>. The paper was noted publicly to lack full " +
"ablations in its first version, so this only shows the <b>scheduling mechanism</b> reduces forgetting " +
"on a controlled stream \u2014 it does <b>not</b> reproduce Hope's language-modeling or needle-in-haystack " +
"results or its claimed wins over Titans / Transformers.";
}
function _tok(s) {
if (s === "live") return null;
if (s === "missing") return "NO-LIVE-DATA";
if (s === "degraded") return "DEGRADED";
if (s === "error") return "OFFLINE";
return "\u2026";
}
function fx(v, d) { return typeof v === "number" ? v.toFixed(d) : "\u2014"; }
function pct(v, d) { return typeof v === "number" ? (v * 100).toFixed(d) + "%" : "\u2014"; }
function _set(id, v) { if (_el[id]) _el[id].textContent = v; }
function _paintOverlay() {
const t = _tok(S.state);
_set("ne-tasks", t || (S.numTasks != null ? String(S.numTasks) : "\u2014"));
_set("ne-clocks", t || (S.kFast != null || S.kMed != null || S.kSlow != null
? (S.kFast != null ? String(S.kFast) : "\u2014") + " / " +
(S.kMed != null ? String(S.kMed) : "\u2014") + " / " +
(S.kSlow != null ? String(S.kSlow) : "\u2014")
: "\u2014"));
_set("ne-gate", t || fx(S.gate, 3));
_set("ne-retbase", t || pct(S.retainBase, 1));
_set("ne-retcms", t || pct(S.retainCms, 1));
_set("ne-delta", t || (S.retDelta != null ? "+" + (S.retDelta * 100).toFixed(1) + "%" : "\u2014"));
_set("ne-writes", t || (S.slowWritesCms != null || S.slowWritesBase != null
? (S.slowWritesCms != null ? String(S.slowWritesCms) : "\u2014") + " / " +
(S.slowWritesBase != null ? String(S.slowWritesBase) : "\u2014")
: "\u2014"));
_set("ne-surprise", t || (S.meanSurprise != null || S.peakSurprise != null
? fx(S.meanSurprise, 3) + " / " + fx(S.peakSurprise, 3)
: "\u2014"));
_set("ne-momentum", t || fx(S.momentum, 2));
// honesty label verbatim — never upgraded
_set("ne-label", t || (S.label || "MODELED"));
if (_plain) _applyPlain();
}
// =============================================================================
// unmount — clean up everything; must not affect other organs
// =============================================================================
export function unmount() {
_polls.forEach((p) => { try { p.stop(); } catch (_) {} }); _polls = [];
try { if (_overlay && _overlay.parentNode) _overlay.parentNode.removeChild(_overlay); } catch (_) {}
try {
if (_group && _stage) {
_group.traverse((o) => {
if (o.geometry && o.geometry.dispose) o.geometry.dispose();
if (o.material) {
const ms = Array.isArray(o.material) ? o.material : [o.material];
ms.forEach((m) => { if (m.dispose) m.dispose(); });
}
});
_stage.scene.remove(_group);
}
} catch (_) {}
_group = _overlay = null;
_floor = null; _rings = []; _tiers = []; _baseMarks = []; _cmsMarks = []; _core = null;
_el = {}; _badge = null; _plain = false; _frameReg = false;
_stage = _THREE = _ctx = null;
S.label = S.numTasks = S.blockLen = S.nTokens = null;
S.kFast = S.kMed = S.kSlow = S.gate = null;
S.taskABase = S.taskACms = null;
S.retainBase = S.retainCms = S.retDelta = S.meanDelta = null;
S.slowWritesCms = S.slowWritesBase = null;
S.meanSurprise = S.peakSurprise = S.momentum = null;
S.curve = null;
S.state = "init";
}
export default { id: ID, title: TITLE, endpoints: [EP], mount, unmount };
|