Unambiguous List Decoding Protocol

• Unambiguous list decoding protocol is a regime where the decoder, although using a list decoder, almost always outputs a singleton containing the transmitted codeword under random error conditions.
• It extends the standard unique decoding radius from δ/2 to δ–ε by leveraging probabilistic analysis and combinatorial bounds on error patterns.
• Algebraic instantiations like Reed–Solomon and AG codes illustrate practical algorithmic implementations, demonstrating the protocol’s real-world applicability.

Searching arXiv for recent and foundational papers related to unambiguous list decoding, random-error list decoding, and related protocol notions. Searching arXiv for "Two Theorems in List Decoding" and complementary work on deterministic RS list decoding and universal polar/list-recovery contexts. An unambiguous list decoding protocol is a decoding regime in which the receiver formally uses a list decoder, yet under the stated channel model the output list is overwhelmingly likely to be a singleton containing only the transmitted codeword. In the most explicit coding-theoretic formulation, this notion is developed for random symbol errors: for any code of relative distance $\delta$ over a sufficiently large alphabet, decoding to radius $\rho n$ with $\rho \le \delta-\varepsilon$ yields, with high probability over the random error values, a list of size $1$, even though worst-case unique decoding is limited to $\rho<\delta/2$ (Rudra et al., 2010).

1. Concept and decoding models

A code $C\subseteq \Sigma^n$ has block length $n$ and relative distance

$$\delta=\frac{1}{n}\min_{c\neq c'\in C}\Delta(c,c'),$$

where $\Delta(\cdot,\cdot)$ is Hamming distance. Unique decoding to radius $\rho n$ means that if there is a codeword within Hamming distance $\rho n$0, it is guaranteed to be unique; in the worst case this is possible only when

$\rho n$1

List decoding to radius $\rho n$2 instead outputs all codewords in the Hamming ball

$\rho n$3

A code is $\rho n$4-list decodable if every such ball contains at most $\rho n$5 codewords (Rudra et al., 2010).

The random-error model that underlies the canonical unambiguous list decoding protocol is more specific than worst-case error correction. The adversary may choose the error locations, but the erroneous symbols themselves are random. If $\rho n$6 is the set of non-error locations, then the error vector $\rho n$7 satisfies $\rho n$8 and $\rho n$9, while on the corrupted coordinates the nonzero symbols are chosen randomly. This is distinct from erasures, where some coordinates are replaced by $\rho \le \delta-\varepsilon$0 and the decoder knows their locations, and also distinct from worst-case symbol errors, where an adversary chooses both locations and replacement symbols (Rudra et al., 2010).

In this usage, “unambiguous” does not mean global unique decodability at radius $\rho \le \delta-\varepsilon$1. It means that the decoder may be a list decoder in principle, but with high probability over the random error pattern the only codeword inside the decoding ball is the transmitted one. The underlying phenomenon is therefore probabilistic unique decoding implemented through list decoding (Rudra et al., 2010). In the broader hierarchy surveyed for list recovery, unique decoding is the special case $\rho \le \delta-\varepsilon$2, list decoding is the special case $\rho \le \delta-\varepsilon$3 of list recovery, and exact unambiguity is therefore the endpoint $\rho \le \delta-\varepsilon$4 of a larger family of soft-information decoding problems (Resch et al., 8 Oct 2025).

2. The random-error singleton theorem

The foundational theorem states that if $\rho \le \delta-\varepsilon$5 is any code over alphabet $\rho \le \delta-\varepsilon$6 of block length $\rho \le \delta-\varepsilon$7 and relative distance $\rho \le \delta-\varepsilon$8, and if $\rho \le \delta-\varepsilon$9 has $1$0, then for sufficiently large alphabet size,

$1$1

and for

$1$2

the following holds: for every codeword $1$3, for all but a $1$4 fraction of error patterns $1$5 with

$1$6

the only codeword within Hamming radius $1$7 of the received word $1$8 is $1$9 (Rudra et al., 2010).

This theorem is stronger than an ordinary bounded-list statement. It does not merely say that the list size is bounded by a constant $\rho<\delta/2$0; it says that with high probability the list collapses to a singleton. The failure probability is exponentially small in $\rho<\delta/2$1. In the proof of part (a), the bad fraction is bounded through inequalities culminating in

$\rho<\delta/2$2

for $\rho<\delta/2$3 and $\rho<\delta/2$4 large enough relative to $\rho<\delta/2$5 (Rudra et al., 2010).

The significance of the theorem is that it exceeds the classical unique-decoding radius by essentially a factor of two. Worst-case unique decoding is limited to

$\rho<\delta/2$6

whereas the random-error theorem gives effective uniqueness for

$\rho<\delta/2$7

Thus the correctable fraction moves from $\rho<\delta/2$8 to $\rho<\delta/2$9, but only in the random-error model and only with high probability (Rudra et al., 2010).

A further variant, Theorem 1(b), allows the search radius to be larger than the actual error fraction. If $C\subseteq \Sigma^n$0 and

$C\subseteq \Sigma^n$1

then for every codeword $C\subseteq \Sigma^n$2, for all but a

$C\subseteq \Sigma^n$3

fraction of error patterns $C\subseteq \Sigma^n$4 with $C\subseteq \Sigma^n$5 and $C\subseteq \Sigma^n$6, the only codeword within radius

$C\subseteq \Sigma^n$7

around $C\subseteq \Sigma^n$8 is $C\subseteq \Sigma^n$9 (Rudra et al., 2010). This enlarges the decoding ball while retaining probabilistic uniqueness.

3. Protocol structure and the disappearance of ambiguity

The corresponding protocol is direct. One fixes any code $n$0 of relative distance $n$1, encodes a message as a codeword $n$2, lets the channel corrupt exactly $n$3 adversarially chosen positions with random nonzero symbols, and receives

$n$4

The decoder then returns the list

$n$5

Under the theorem’s assumptions,

$n$6

The decoder is therefore formally a list decoder, but operationally an unambiguous decoder with overwhelming probability (Rudra et al., 2010).

The result is information-theoretic and combinatorial rather than algorithmic for arbitrary code families. Its direct guarantees rely on the code’s relative distance $n$7, the random-error condition $n$8, the alphabet-size requirement

$n$9

and the asymptotic regime of large $$\delta=\frac{1}{n}\min_{c\neq c'\in C}\Delta(c,c'),$$0 (Rudra et al., 2010). A plausible implication is that practical deployment requires a code family that already has an efficient list decoder or some other structured search procedure.

The proof mechanism is a counting argument on “bad” error patterns. Fixing the transmitted codeword $$\delta=\frac{1}{n}\min_{c\neq c'\in C}\Delta(c,c'),$$1, an error pattern is $$\delta=\frac{1}{n}\min_{c\neq c'\in C}\Delta(c,c'),$$2-bad if there exists another codeword $$\delta=\frac{1}{n}\min_{c\neq c'\in C}\Delta(c,c'),$$3 with

$$\delta=\frac{1}{n}\min_{c\neq c'\in C}\Delta(c,c'),$$4

The proof partitions the agreement set into agreement on non-error positions and agreement on error positions, and uses the relative-distance condition to show that once more than $$\delta=\frac{1}{n}\min_{c\neq c'\in C}\Delta(c,c'),$$5 positions of a codeword are fixed there can be at most one codeword consistent with them. This yields a bound of the form

$$\delta=\frac{1}{n}\min_{c\neq c'\in C}\Delta(c,c'),$$6

with $$\delta=\frac{1}{n}\min_{c\neq c'\in C}\Delta(c,c'),$$7 bounded crudely by $$\delta=\frac{1}{n}\min_{c\neq c'\in C}\Delta(c,c'),$$8 in part (a), so the random-symbol coincidence needed to create ambiguity is exponentially unlikely (Rudra et al., 2010).

The large-alphabet assumption is essential in the theorem’s universal form. The paper shows that if

$$\delta=\frac{1}{n}\min_{c\neq c'\in C}\Delta(c,c'),$$9

then the statement of part (a) is false in general (Rudra et al., 2010). This sharply separates the random-error singleton phenomenon from a blanket statement over all alphabets.

4. Algorithmic realizations in algebraic codes

The paper gives a concrete Reed–Solomon instantiation. For Reed–Solomon codes of dimension $\Delta(\cdot,\cdot)$0, block length $\Delta(\cdot,\cdot)$1, over $\Delta(\cdot,\cdot)$2 with

$\Delta(\cdot,\cdot)$3

it proves that for at least $\Delta(\cdot,\cdot)$4 fraction of error patterns $\Delta(\cdot,\cdot)$5 of Hamming weight at most $\Delta(\cdot,\cdot)$6, and any codeword $\Delta(\cdot,\cdot)$7, the only codeword that agrees in at least $\Delta(\cdot,\cdot)$8 positions with $\Delta(\cdot,\cdot)$9 is $\rho n$0. The resulting algorithm enumerates all subsets $\rho n$1 of size $\rho n$2, checks whether the restriction lies in the projected Reed–Solomon code, reconstructs the unique candidate extension when it exists, and runs in

$\rho n$3

time (Rudra et al., 2010). This is an explicit unambiguous list decoding protocol in the random-error sense because its correctness rests on the fact that, with high probability, only the transmitted codeword satisfies the large-radius agreement condition.

A later deterministic Reed–Solomon result takes a different algorithmic route. There is a deterministic algorithm that, for every finite field $\rho n$4 and parameters $\rho n$5, runs in time $\rho n$6 and list decodes Reed–Solomon codes from agreement greater than

$\rho n$7

It outputs all degree-$\rho n$8 polynomials $\rho n$9 satisfying the threshold, and therefore yields a deterministic unambiguous protocol whenever external conditions ensure that the list size is $\rho n$00 or $\rho n$01. The paper does not itself prove a new singleton theorem at the Johnson threshold; uniqueness is checked afterward by inspecting the output list (Chatterjee et al., 7 Nov 2025).

For general one-point algebraic-geometry codes, a Gröbner-basis decoder based on voting supports both list decoding and an explicit unique-decoding regime. It applies to ordinary one-point AG codes $\rho n$02 and Feng–Rao improved codes $\rho n$03, branches whenever multiple voting-consistent symbol candidates appear, and becomes effectively unique whenever

$\rho n$04

Outside that regime it remains a genuine list decoder, but its voting rule and termination criteria act as ambiguity-reducing devices that may certify a single candidate branch (Matsumoto et al., 2012).

5. Other protocol architectures based on list generation and disambiguation

The random-error singleton theorem is not the only way to approach unambiguity. A different architecture first generates a controlled list and then resolves it using auxiliary information. In nested lattice coding for relay channels, a decoder observes

$\rho n$05

and outputs a list of size

$\rho n$06

that contains the transmitted lattice point with high probability. The list is then disambiguated using a bin index or side information from another block or node. The decoder is intentionally ambiguous by itself, but the overall network protocol becomes effectively unambiguous after list intersection (Song et al., 2010). This suggests a two-stage notion of unambiguity: ambiguity may be suppressed either by the noise model itself or by combining a list with independent side information.

In ARQ systems for linear codes, list decoding is used to correct undetected errors after hard decision and retransmission. For each accepted received codeword $\rho n$07, the candidate list is

$\rho n$08

where $\rho n$09 is the code minimum distance, and a Context List Decoding Algorithm then selects a single final codeword sequence by dynamic programming under a Markov context model. The final output is unique, but the correctness guarantee is probabilistic rather than combinatorial; the ambiguity is resolved by context statistics rather than by a singleton theorem for every realization (Wang et al., 2018).

Universal polar coding provides yet another adjacent pattern. A successive-cancellation list decoder for universal polar codes keeps $\rho n$10 most likely paths and can be implemented in $\rho n$11 space and $\rho n$12 time for the full universal transform. The paper does not specify a final unambiguous selector beyond the usual notion of keeping the most likely paths, and CRC-like mechanisms are mentioned only peripherally. Here list maintenance is explicit, while true unambiguity must be supplied by a separate policy or external criterion (Shuval et al., 2020).

6. Limits, neighboring notions, and contrasting uses of ambiguity

A central limitation is that small-list or capacity-achieving list decoding is not the same as unambiguous list decoding. Randomly punctured Reed–Solomon codes over fields as small as

$\rho n$13

can be $\rho n$14 average-radius list-decodable with high probability, and the list size $\rho n$15 is the optimal order near capacity. This is bounded ambiguity, not singleton collapse, and the result is combinatorial rather than algorithmic (Alrabiah et al., 2023). The broader survey of list recoverable codes makes the same point in a more general language: exact unambiguity is $\rho n$16, whereas near-capacity behavior typically yields list sizes such as $\rho n$17 rather than $\rho n$18 (Resch et al., 8 Oct 2025).

The rate–ambiguity tradeoff can also be expressed abstractly. In ratio list decoding, the decisive quantity is

$\rho n$19

and the ratio-capacity, defined as the supremum of achievable normalized logarithms of $\rho n$20, is equal to Shannon capacity $\rho n$21. Since $\rho n$22 corresponds to unique decoding, this framework identifies strict unambiguity as the endpoint of a broader continuum in which ambiguity is quantified by list size rather than eliminated outright (Somekh-Baruch, 2018).

Several nearby literatures reverse the objective entirely and treat ambiguity as a resource. Secure list decoding requires that the transmitted message be contained in the list while the receiver cannot uniquely identify it and the sender cannot control another list element (Hayashi, 2019). List privacy amplification in quantum key distribution likewise extracts a list of $\rho n$23 candidate keys such that one hidden-index key is truly secret, with achievable key length

$\rho n$24

so the gain comes from preserving and hiding ambiguity rather than eliminating it (Kulkarni, 18 Mar 2026). These constructions are therefore not unambiguous list decoding protocols in the coding-theoretic sense; they are controlled-ambiguity protocols.

The most precise coding-theoretic use of the term remains the random-error singleton phenomenon. In that setting, unambiguity means that list decoding is necessary in principle because $\rho n$25 may exceed $\rho n$26, but under random symbol corruptions the list is overwhelmingly likely to be $\rho n$27. The concept is therefore best viewed as probabilistic unique decoding via list decoding, with efficient realization depending on whether the chosen code family supplies a complete list decoder or an equivalent structured search method (Rudra et al., 2010).