diff --git a/README.md b/README.md
index b978638..7898d54 100644
--- a/README.md
+++ b/README.md
@@ -137,6 +137,36 @@ where `x̂ = γ · LN(x) + γₓ · x` (scaled LayerNorm, `γ` init `10⁻⁶`,
 
 where `A ∈ ℝ^{r×d}`, `B ∈ ℝ^{d×r}`, `r = 4`
 
+### Pair formation and negative sampling
+
+The dataset provides only **positive pairs** — each entry maps one drone image to its
+matching satellite crop(s). Negative pairs are **not** stored explicitly; instead, they
+are constructed automatically inside each training batch via the InfoNCE loss:
+
+```
+Batch (B = 8):
+  drone_0 ↔ sat_0  ← positive (from dataset)
+  drone_1 ↔ sat_1  ← positive
+  ...
+  drone_7 ↔ sat_7  ← positive
+
+Similarity matrix B×B:
+       sat_0  sat_1  ...  sat_7
+  q_0 [ pos    neg   ...   neg ]   ← CE target = 0
+  q_1 [ neg    pos   ...   neg ]   ← CE target = 1
+  ...
+  q_7 [ neg    neg   ...   pos ]   ← CE target = 7
+```
+
+- **Positives:** diagonal `sim[i, i]` — correct drone-satellite pair
+- **Negatives:** off-diagonal `sim[i, j≠i]` — other satellite images in the batch (in-batch negatives)
+- At `batch_size=8`: 1 positive + 7 in-batch negatives per query
+- No hard-negative mining — negatives are random within the batch
+- Larger batch → more negatives → harder contrastive task
+
+Each drone image may have multiple satellite candidates (with distance-based weights).
+At training time, one satellite crop is sampled per drone (weighted if semi-positive).
+
 ### Loss function
 
 Symmetric InfoNCE with learnable temperature (CLIP-style `logit_scale`):