diff --git a/misc/prio3_multiproof_robustness.py b/poc/plot_prio3_multiproof_robustness.py
similarity index 50%
rename from misc/prio3_multiproof_robustness.py
rename to poc/plot_prio3_multiproof_robustness.py
index 0c44c56c..77a3e62f 100644
--- a/misc/prio3_multiproof_robustness.py
+++ b/poc/plot_prio3_multiproof_robustness.py
@@ -3,8 +3,8 @@
 import matplotlib.pyplot as plt
 import math
 
-FIELD128_MODULUS = 2**66 * 4611686018427387897 + 1  # Field128.MODULUS
-FIELD64_MODULUS = 2**32 * 4294967295 + 1            # Field64.MODULUS
+from field import Field64, Field128
+from vdaf_prio3 import Prio3SumVec
 
 NUM_REPORTS = 1000000000
 
@@ -22,10 +22,10 @@ def soundness(gadget_calls, gadget_degree, field_size):
     return gadget_calls * gadget_degree / (field_size - gadget_calls)
 
 
-def robustness(epsilon, ro_queries, prep_queries, num_proofs):
+def robustness(epsilon, ro_queries, prep_queries, num_proofs, seed_bits):
     '''
-    ia.cr/2023/130, Theorem 1, ignoring negligible terms and assuming the bound
-    can be modified by raising `epsilon` to the power of the number of FLPs.
+    ia.cr/2023/130, Theorem 1, assuming the bound can be modified by raising
+    `epsilon` to the power of the number of FLPs.
 
     epsilon - soundness of the base FLP
 
@@ -35,8 +35,11 @@ def robustness(epsilon, ro_queries, prep_queries, num_proofs):
     prep_queries - number of online attempts, a proxy for the batch size
 
     num_proofs - number of FLPs
+
+    seed_bits - the size of the XOF seed in bits
     '''
-    return (ro_queries + prep_queries) * epsilon**num_proofs
+    return (ro_queries + prep_queries) * epsilon**num_proofs + \
+           (ro_queries + prep_queries**2) / 2**(seed_bits - 1)
 
 
 def sum_vec(field_size, num_proofs, length):
@@ -45,50 +48,44 @@ def sum_vec(field_size, num_proofs, length):
     accepting one report in a batch of NUM_REPORTS. Assuming the asymptotically
     optimal chunk length.
     '''
+    chunk_length = max(1, int(math.sqrt(length)))
+    vdaf = Prio3SumVec.with_params(length, 1, chunk_length)
+    gadget_calls = vdaf.Flp.Valid.GADGET_CALLS[0]
+
+    base_flp_soundness = soundness(gadget_calls, 2, field_size)
 
-    # Table 11
-    def gadget_calls(length, bits, chunk_length):
-        return (length * bits + chunk_length - 1) // chunk_length
+    # SumVec interprets the inner Mul-gadget outputs as coefficients of a
+    # polynomial and evaluates the polynomial at a random point. If a gadget
+    # output is non-zero, then the output is non-zero except with this
+    # probability. This is bounded by the number of roots of the polynomial.
+    circuit_soundness = length / field_size
 
     return robustness(
-        soundness(
-            gadget_calls(length, 1, max(1, math.sqrt(length))),
-            2,
-            field_size,
-        ),
-        2**80,       # ro_queries
-        NUM_REPORTS,  # prep_queries
-        num_proofs,  # num_proofs
+        base_flp_soundness + circuit_soundness,  # ia.cr/2019/188, Theorem 5.3
+        2**80,
+        NUM_REPORTS,
+        num_proofs,
+        vdaf.Xof.SEED_SIZE * 8,
     )
 
-print(math.log2(sum_vec(FIELD128_MODULUS, 1, 1000000)))
+print(math.log2(sum_vec(Field128.MODULUS, 1, 1000000)))
 
-lengths = range(0, 1000000, 100)
+lengths = range(100, 10**5, 100)
 plt.plot(
     lengths,
-    [sum_vec(FIELD128_MODULUS, 1, length) for length in lengths],
+    [sum_vec(Field128.MODULUS, 1, length) for length in lengths],
     label='Field128/1',
 )
 plt.plot(
     lengths,
-    [sum_vec(FIELD64_MODULUS, 1, length) for length in lengths],
-    label='Field64/1',
-)
-plt.plot(
-    lengths,
-    [sum_vec(FIELD64_MODULUS, 2, length) for length in lengths],
+    [sum_vec(Field64.MODULUS, 2, length) for length in lengths],
     label='Field64/2',
 )
 plt.plot(
     lengths,
-    [sum_vec(FIELD64_MODULUS, 3, length) for length in lengths],
+    [sum_vec(Field64.MODULUS, 3, length) for length in lengths],
     label='Field64/3',
 )
-plt.plot(
-    lengths,
-    [sum_vec(FIELD64_MODULUS, 4, length) for length in lengths],
-    label='Field64/4',
-)
 
 plt.xscale('log', base=10)
 plt.yscale('log', base=2)