Update the README for the current figures and results

Fig. 3 is now the 10 dB aggregate with the token-domain bars, the
non-MAML reference carries its figure label, the load-sweep script is
listed, and the probe reference value matches the reported runs.

README.md

@@ -1,27 +1,30 @@
-# Semantic Multiplexing Gain in Wireless Systems via Expanded Embeddings: A BERT Case Study
+# Semantic Multiplexing Gain in Wireless Systems via Expanded Embeddings: A BERT Case Study
 
 Code, stored results, and supplementary material for the IEEE
 Communications Letters submission by Ki-Ho Lee, Hyun-Ho Choi, and
 Jung-Ryun Lee.
 
 Multiple users share one expanded embedding block of dimension
-`d_s = K * d_b`: each user's frozen BERT sentence embedding is projected
+`d_s = K * d_b`. Each user's frozen BERT sentence embedding is projected
 into the shared space, superimposed through learnable masks, and
-demultiplexed by user-wise attention. All reported transceivers are
-trained with SNR-aware MAML; the training without MAML of the earlier
-JSAC paper is included as a prior-art reference.
+demultiplexed by user-wise attention. Every transceiver reported in the
+letter, including the conventional orthogonal scheme, is trained with
+SNR-aware MAML; the same architecture trained without MAML, as in the
+earlier JSAC paper, is included as a prior-art reference and is labelled
+"Training w/o MAML [5]" in the figures.
 
 ## Files
 
 | File | Purpose |
 |---|---|
 | `bert_semcom.py` | Shared library: BERT extractor, transceiver model, channel, MAML helpers |
-| `cl_experiments.py` | Held-out split, joint-trained configurations, ToDMA token-domain benchmark, linear probe, latency |
-| `cl_maml_all.py` | SNR-aware MAML training for every reported configuration (including the conventional orthogonal scheme) |
-| `cl_maml_extra.py` | MAML K sweep (K = 1, 2, 8) and DistilBERT replication |
+| `cl_experiments.py` | Held-out split, configurations trained without MAML, token-domain (ToDMA) comparison, linear probe, latency |
+| `cl_maml_all.py` | SNR-aware MAML training for every reported configuration, including the conventional orthogonal scheme |
+| `cl_maml_extra.py` | MAML expansion-factor sweep (K = 1, 2, 8) and DistilBERT replication |
+| `cl_todma_load.py` | Evaluates the token-domain scheme at U = 1, 2, 3, 5, 6 for the load sweep of Fig. 3 |
 | `replot_cl.py` | Regenerates Figs. 2 and 3 of the letter from the stored JSON results |
 | `probe_vs_cosine.py` | Supplementary probe-accuracy-versus-cosine-similarity analysis |
-| `fig_cl/*.json`, `fig_cl/*.csv` | Stored raw results behind every figure and quoted number |
+| `fig_cl/*.json`, `fig_cl/*.csv` | Stored raw results behind every figure and every number quoted in the letter |
 
 ## Reproducing
 
@@ -30,52 +33,72 @@ Requirements: Python 3.10+, PyTorch (CUDA), `transformers`, `datasets`,
 (`fancyzhx/ag_news` fallback included).
 
 ```bash
-python cl_experiments.py --save-dir fig_cl   # joint runs + ToDMA benchmark (~3 h on a laptop GPU)
+python cl_experiments.py --save-dir fig_cl   # runs without MAML + token-domain comparison (~3 h on a laptop GPU)
 python cl_maml_all.py    --save-dir fig_cl   # MAML runs (~9 h)
 python cl_maml_extra.py  --save-dir fig_cl   # MAML K sweep + DistilBERT (~6 h)
-python replot_cl.py                          # Figs. 2 and 3 from stored results
+python cl_todma_load.py  --save-dir fig_cl   # token-domain load sweep (evaluation only)
+python replot_cl.py                          # Figs. 2 and 3 from the stored results
 python probe_vs_cosine.py                    # supplementary analysis below
 ```
 
 All experiments fix their random seeds (training seed 42, evaluation
-seed 123, ToDMA seed 7) and evaluate on a held-out test split of 2,000
-AG News sentences disjoint from the 8,000-sentence training pool.
-`replot_cl.py` and `probe_vs_cosine.py` read only the stored results,
-so every figure is regenerable without rerunning the experiments.
+seed 123, token-domain seed 7) and evaluate on a held-out test split of
+2,000 AG News sentences disjoint from the 8,000-sentence training pool.
+The centering mean, the transceiver parameters, and the linear probe are
+fitted on the training pool only. `replot_cl.py` and
+`probe_vs_cosine.py` read only the stored results, so every figure is
+regenerable without rerunning the experiments.
 
 ## Figures of the letter
 
-**Fig. 2 - per-user cosine similarity vs. SNR** (proposed scheme for
-U = 1..4 at K = 4, the conventional orthogonal scheme, the
-matched-budget schemes, and the joint training of the earlier JSAC
-paper, all on the held-out test set):
+**Fig. 2 - per-user cosine similarity vs. SNR.** The proposed scheme for
+U = 1..4 at K = 4, the conventional orthogonal scheme, the two
+matched-budget comparison schemes (random-projection mask and the
+token-domain scheme at two slot/codeword splits), and the same
+architecture trained without MAML:
 
 ![Fig. 2](fig_cl/cl_fig_mux.png)
 
-**Fig. 3 - aggregate fidelity across load** (SNR-aware MAML vs. joint
-training at 20 dB, with the fully loaded orthogonal reference):
+**Fig. 3 - aggregate fidelity across load at 10 dB.** Three bars per
+load point (proposed, training without MAML, and the token-domain
+scheme at 24x128), the conventional single-user block on the left, and
+the fully loaded orthogonal aggregate as the dash-dotted reference:
 
 ![Fig. 3](fig_cl/cl_fig_agg.png)
 
+At full load the shared block stays within 1% of the fully loaded
+orthogonal aggregate over the same 3072 channel uses (3.59 against
+3.62), and under overload it reaches 1.45 times that aggregate at
+U = 6.
+
 ## Supplementary: probe accuracy vs. cosine similarity
 
 The letter measures semantic fidelity by the cosine similarity of the
 recovered embeddings and corroborates it with a downstream perception
-metric: the AG News topic accuracy of a linear probe trained on clean
-training-pool embeddings and applied to the recovered test embeddings
-(clean reference about 0.855, sampling error about +/-0.01).
+metric, namely the AG News topic accuracy of a linear probe fitted on
+the training pool and applied to the recovered test embeddings. The
+noiseless reference is about 0.854 and the sampling error is about
++/-0.01.
 
 Across 7 schemes x 7 SNRs (49 operating points), probe accuracy tracks
 cosine similarity with a Pearson correlation of **r = 0.903**:
 
 ![Probe accuracy vs. cosine similarity](fig_cl/probe_vs_cosine.png)
 
-Two readings follow. First, the low-SNR advantage of the analog
-embedding schemes over the token-domain scheme appears in both metrics
-(for example 0.848 vs. 0.772 in CosSim and 0.805 vs. 0.762 in accuracy
-at 5 dB). Second, schemes within about 0.01 of each other in CosSim
-differ in accuracy only on the order of the sampling error, so the
+Two readings follow. First, the low-SNR advantage of the embedding
+schemes over the token-domain scheme appears in both metrics, for
+example 0.848 against 0.772 in cosine similarity and 0.805 against
+0.762 in accuracy at 5 dB. Second, schemes within about 0.01 of each
+other in cosine similarity differ in accuracy by at most 0.023, which
+is within about two standard errors of the sampling noise, so the
 cosine metric used throughout the letter is consistent with downstream
 perception on this task.
 
+## Citation
 
+The letter is under review. Until it appears, please cite this
+repository together with the earlier shared-embedding paper:
+
+> K.-H. Lee, H.-H. Choi, and J.-R. Lee, "Transformer-based shared
+> embedding for multiple access in semantic communications," *IEEE J.
+> Sel. Areas Commun.*, vol. 44, pp. 2622-2637, 2026.