diff --git a/README.md b/README.md
index 673ac99..63afab6 100644
--- a/README.md
+++ b/README.md
@@ -1020,6 +1020,16 @@ process uses its `LOCAL_RANK`.  DDP intentionally rejects
 `--online-current-tail`; the online producer-consumer path remains a separate
 benchmark.
 
+To continue from a completed run, pass both its `adapter/` directory through
+`--adapter-checkpoint` and its `turn_head.pt` through `--head-checkpoint` while
+writing to a new output directory.  This is weights-only continuation: the
+adapter and head weights resume, but AdamW moments and the cosine scheduler are
+reinitialized.  Preserve the previous global batch when changing GPU count;
+for example, change 8 GPUs with batch 2 and accumulation 1 to 4 GPUs with batch
+2 and accumulation 2.  Use zero warmup and a reduced learning rate for a short
+post-hoc continuation because a one-epoch cosine schedule has already reached
+zero.
+
 For a speed-only alternative, `train_omni_turn_lora.py` supports a two-GPU
 producer-consumer benchmark.  The producer GPU runs the frozen AV towers and
 places current-tail embeddings in a bounded CPU queue while the consumer GPU
@@ -1112,6 +1122,43 @@ semantics. Run the following controlled experiments:
 The frozen-Omni head is the cleanest representation test. LoRA can be reported
 as a second-stage ceiling, not as the only result.
 
+## Candidate improvement: explicit working and long memory
+
+Motivated by EgoSpeak's separation of long-term and working memory, test an
+explicit two-timescale representation instead of compressing the entire causal
+window into one undifferentiated final Thinker state.  At prediction time `t`,
+align long memory to fixed global 2-second boundaries and let working memory
+slide with the current prediction time:
+
+```text
+b = floor(t / 2 s) * 2 s
+long memory:     complete 2-second chunks ending at b, ..., [b-4, b-2], [b-2, b]
+working memory:  the sliding recent interval [t-2, t], updated every 0.2 s
+```
+
+For example, at `t = 20.4 s`, long memory contains the fixed chunks
+`[16,18]`, `[18,20]`, and any earlier retained chunks, while the working
+partial is `[18.4,20.4]`.  The overlap between `[18,20]` and `[18.4,20.4]` is
+intentional: long memory provides a stable completed-chunk representation,
+whereas working memory preserves the exact high-resolution context ending at
+the current prediction time.  Cache one representation per completed
+long-memory chunk, recompute the working-memory representation at each 5 Hz
+prediction step, and fuse the two branches with a small gated/attention
+pooling module before the same `10 x 3` prediction head.  Keep the Omni
+backbone frozen in the first test so that any gain can be attributed to memory
+organization rather than additional backbone capacity.
+
+This is a candidate score-improvement idea, not yet an experimental result.
+The key hypothesis is that a dedicated recent branch preserves local prosody,
+speaker-change, and utterance-completion cues while long memory supplies
+conversation context without diluting them in one final state.  Evaluate it
+against matched-compute baselines: (1) the current single-state 8-second model,
+(2) working-memory-only `[t-2,t]`, and (3) long+working memory with the same
+total causal context.  Report parameter count, throughput, and per-horizon and
+per-class AP over identical seeds.  A useful result requires the fused model to
+beat both controls consistently; otherwise a gain could merely come from the
+extra fusion parameters or the stronger recent-window shortcut.
+
 ## Single-window feature inspection
 
 `qwen_omni_features.py` is useful for inspecting a short window or layer, but