Fine-tuning ORBGRAND with Very Few Channel Soft Values

Huarui Yin; Li Wan; Wenyi Zhang

REVIEW 3 major objections 4 minor 1 cited by

Fine-tuning ORBGRAND with Very Few Channel Soft Values

T0 review · 3 major / 4 minor · reviewed 2026-08-06 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read Using the exact value of just one log-likelihood ratio, fine-tuned ORBGRAND approaches maximum-likelihood block error rate at nearly ORBGRAND's complexity.

desk verdict Fine-tuning ORBGRAND with one or two exact LLRs is a genuinely new idea with a clear BLER gain, but the key position formula (18) is an unquantified density approximation and the evaluation is single-code, so the result is promising rather than proven. read the letter →

arxiv 2507.08696 v1 pith:YL63SNVK submitted 2025-07-11 cs.IT math.IT

Li Wan , Huarui Yin , Wenyi Zhang This is my paper

classification cs.ITmath.IT MSC 94B3594A1505A17

keywords GRANDORBGRANDsoft-decisiondecodingorderedreliabilitybitsintegerpartitionsreversepairsmaximum-likelihoodshortblock-lengthcodes

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This paper tries to show that the gap between practical GRAND decoders and maximum-likelihood decoding can be closed using almost no extra channel information. It proposes a fine-tuning step for ORBGRAND: replace the rank-based weight of one or two carefully chosen bits with the bit's exact log-likelihood magnitude, then reorder the error-pattern test list accordingly. The choice of which bits to fine-tune is made by counting reverse pairs, pairs of error patterns tested in the wrong probability order, and by estimating that count with the asymptotic theory of integer partitions. On a BCH(127,113) code, the fine-tuned decoder with D=1 is reported to beat all ORB-type GRAND baselines and to land within 0.05 dB of SGRAND, at an added cost of about one extra operation per test. If this holds, near-ML decoding for short high-rate codes no longer requires the expense of full soft-output sorting.

What carries the argument

The central object is the reverse-pair metric $\eta$, the fraction of pairs among the first $T$ tested error patterns whose soft weights are inverted relative to the test order; SGRAND has $\eta = 0$, ordinary ORBGRAND has positive $\eta$. The engine of the method is the asymptotic estimate (15) of $O(m)$, the number of error patterns with ORB-weight at most $m$, based on the classical formula for non-repeating partitions cited as Lemma 1, and the resulting estimate (16) of the reverse-pair count as a sum over partitions of the $D$ selected positions of $p_u(T)p_v(T)|\delta(u,v)|$ times an integral over $\widetilde{O}'(m)$. These lead to the position-adjustment formula (18) and the simplified selection rule (19), which for $D=1$ is $d_1^* = \arg\max_{d_1} p_0(T)(1-p_0(T))|\ell_{d_1} - \gamma_{r_{d_1}}|$.

What would settle it

Simulate fine-tuned CDF-ORBGRAND on a different code, such as BCH(63,51) or BCH(255,239), at $E_b/N_0 = 5$ dB with D=1; if the positions selected by (19) do not maximize the exact reverse-pair count (12), or the BLER gap to SGRAND is markedly larger than the 0.05 dB reported for BCH(127,113), the central claim would fail.

Watch

Extended reading notes

Core claim

The paper's central claim is that exact channel soft values are only needed at a very small set $D$ of positions for an ORB-type GRAND decoder to nearly match maximum-likelihood decoding. The positions are chosen to maximize $I$, the number of reverse pairs between tested error patterns under the modified weight $\widetilde{\zeta}(e) = \Gamma(e) + \delta(u)$, where $\delta(u) = \sum_{i \in D}(\ell_i - \gamma_{r_i}) e_i$ depends only on the selected positions. The paper estimates $I$ using the partition-based approximation (15) for $O(m)$, the number of error patterns with $\Gamma \le m$, which yields the closed-form estimate (16) and, after an empirical simplification, the selection rule (19). For $D=1$ that rule reduces to choosing $d_1$ to maximize $p_0(T)(1-p_0(T))|\ell_{d_1} - \gamma_{r_{d_1}}|$. Using formula (18) to reorder error patterns on the fly, the paper reports BLER within 0.05 dB of SGRAND on a BCH(127,113) code.

Load-bearing premise

The load-bearing premise is the empirical observation, shown only at one signal-to-noise ratio with D=1, that the benefit of a fine-tuning position is insensitive to the magnitude $|\ell_i - \gamma_{r_i}|$; if that insensitivity fails for other code lengths, rates, SNRs, or D=2, the selection rule (19) will not pick the positions that most improve ordering.

Editorial extensions

If this is right

Fine-tuned CDF-ORBGRAND with D=1 achieves lower BLER than ORBGRAND, CDF-ORBGRAND, and RS-ORBGRAND, and comes within 0.05 dB of SGRAND.
The average number of tests and the average decoding delay are no larger than those of untuned CDF-ORBGRAND, because the reordering reduces the number of guesses needed.
The reverse-pair metric $\eta$ drops from 0.233 (CDF-ORBGRAND) to 0.211 (fine-tuned, D=1) at 7 dB, showing the test list is closer to maximum-likelihood order.
The added complexity is negligible: the while-loop adjustment costs O(1) per test on average and O(T) in the worst case, in contrast to SGRAND's exponential generation cost per test.
The framework applies to any ORB-type GRAND whose basis list satisfies condition (3); the only algorithm-specific inputs are $\gamma$ and a fitted estimate of $O'(m)$.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

Not tested in the paper: whether D=1 remains sufficient at longer block lengths, where T=10^4 covers a much smaller fraction of the $2^N$ pattern space; the partition estimate suggests the reverse-pair structure changes with N, so the 0.05 dB gap may widen.
Not tested in the paper: the same machinery could transfer to non-Gaussian channels by recomputing the LLR distribution and refitting the $O'(m)$ estimate, since the rank-based weight $\gamma$ would change.
One could also invert the selection criterion: instead of maximizing $p_0(T)(1-p_0(T))|\ell - \gamma_r|$, choose positions where the flip probability is most balanced across subsets, which would reduce dependence on the empirical insensitivity of (16).

Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Desk Editor's Note

read the letter

New thing here: a fine-tuning rule that spends one or two exact LLR magnitudes to reorder the ORBGRAND error-pattern list, chosen via a reverse-pair metric and an integer-partition estimate of pattern positions. That is a legitimate new idea in the ORB-type GRAND family, not present in the cited work.

What's good: the BLER curves on BCH(127,113) speak clearly. With D=1 the decoder lands within about 0.05 dB of SGRAND and beats RS-ORBGRAND, at an average overhead of O(1) extra tests. The partition-based estimate of O(m) is remarkably accurate in Figure 2, and the reverse-pair count estimate in Figure 3 is within 1% of exact. The authors also label the derivation as heuristic in Section V, so they aren't overselling the theory.

The soft spot is exactly the one the stress-test note flags. Equation (18) is load-bearing: every fine-tuned position in Algorithm 1 comes from it. But p_v is defined in (17) as a fraction among the first T patterns, and in (18) it is evaluated at an arbitrary threshold count O(Γ(e)+δ(u,v)). The proof substitutes a subset count by p_v times O(.) with no stated error bound. That is a density approximation with no direct validation. Figure 4(a) tests only the insensitivity of the integral in (16) to |δ| for D=1 at 7 dB; it doesn't test the accuracy of (18). Because position errors and selection errors can compensate, the near-SGRAND D=1 result isn't established at the level of a theorem. I also wish the evaluation weren't a single code with no error bars or artifact release; that makes it hard to know whether the gain transfers to other lengths and rates.

None of this is fatal. The empirical case is real, the method is simple, and the authors are honest about the heuristic nature. What's missing is either a rigorous bound on (18) or a much broader simulation study. The paper deserves a serious referee; it is a solid within-subfield improvement that moves the ORBGRAND-vs-SGRAND gap in the right direction. I would not cite it in my own work yet as a proven technique, but I would certainly cite it as a promising approach and read a revised version carefully. Who is this for: people working on GRAND variants or short-block URLLC decoders. Send it to review, with instructions to push on the position-formula approximation and the single-code evaluation.

Referee Report

3 major / 4 minor

Summary. The paper proposes a fine-tuning method for ORB-type GRAND decoding that uses a very small number of exact channel soft values, D=1 or D=2, to reorder the error-pattern test sequence. The ordering is guided by a reverse-pair 'well-orderedness' metric, and the paper estimates the metric using asymptotics of integer partitions. The main theoretical objects are Eq. (16), an approximate expression for the number of reverse pairs, and Eq. (18), a formula for the rearranged position of each error pattern. On BCH(127,113) over AWGN, the reported BLER curves show that fine-tuned CDF-ORBGRAND with D=1 outperforms the listed ORB-type GRAND baselines and approaches SGRAND, with a small increase in average decoding delay.

Significance. If the central finding holds, the method is a practically relevant bridge between low-complexity ORB-type GRAND and ML-approaching SGRAND: it offers most of the BLER gain with only one or two exact reliability values and with empirically negligible overhead. The paper is honest about the heuristic nature of the development: the conclusion explicitly labels the derivation as heuristic, and the numerical sections verify the estimation accuracy of the partition-based count. The empirical checks of O(m) and of the reverse-pair count I against exact values are a strength, as is the inclusion of several ORB-type baselines and SGRAND in the BLER comparison. However, the main algorithmic step relies on an unvalidated density approximation in Eq. (18), and the empirical evidence covers only one code and one channel model.

major comments (3)

[Section III-D, Theorem 2, Eq. (18)] The position formula in (18) is load-bearing because Algorithm 1, line 7, uses it to set A(t_FT), yet its proof is not valid as written. The quantity p_v(.) is defined in (17) only as an empirical fraction over the first T patterns, p_v(T), not as a function of an arbitrary threshold O(Gamma(e)+delta(u,v)). The proof silently replaces the exact subset count #{e(t) in E_v : Gamma(e(t)) <= Gamma(e)+delta(u,v)} by p_v(O(...)) O(...), which is a density approximation with no stated error bound and no direct validation. Since an inaccurate position estimate changes which error pattern is tested at each step, the near-SGRAND D=1 BLER result in Figure 5 could in principle arise from a compensation between errors in (18) and in the selection rule (19). Please either derive (18) with a controlled approximation or, more practically, validate it directly: for simulated received vectors, compute the position predicted by (18) and compare it with the true position under the zeta-tilde criterion, reporting mean and quantile errors and the resulting effect on BLER.
[Section III-D, Eq. (19), Figure 4(a)] The simplification from (16) to (19) is justified only by the empirical observation that the integral in (16) is insensitive to |delta(u,v)| for D=1 at Eb/N0=7 dB for CDF-ORBGRAND. The D=2 selection rule and the claimed D=2 performance in Figure 5 rely on the same insensitivity without any check. Please provide the analogous insensitivity curves for D=2 and for at least one other operating point and code parameter. If the insensitivity fails there, the selected fine-tuning positions will not maximize the reverse-pair count and the BLER gains may not transfer.
[Section IV-B, Figure 5] The central claim that 'we outperform all ORB-type GRAND algorithms and come very close to SGRAND' is supported only for BCH(127,113) with T=10^4 and a single modulation and channel model. This is a load-bearing generalization. Adding even one more code with a different length or rate, or a CRC-aided polar code, and reporting the same BLER comparison would materially strengthen the claim; as it stands, the near-SGRAND D=1 result is a single empirical point.

minor comments (4)

[Section III-C, Eq. (16)] The sentence preceding (16) says the detailed derivation steps are omitted; since (16) is used to justify the selection rule, please state explicitly which approximations are made (replacing sums by integrals, treating p_v as a smooth density, and ignoring boundary terms) so a reader can assess the conditions under which the approximation should hold.
[Section III-D, Algorithm 1] In the line 'u <- bin2dec([ed*_1(idx), ..., ed*_D(idx)])', the bit-ordering convention for dec2bin/bin2dec is not specified. Please clarify whether the first entry is the least significant bit, since the definition of delta(u) in Proposition 1 depends on that convention.
[Section IV-B, Figure 4(a)] The caption and text say 'the i is considered to be uniformly distributed', which is unclear; it should be stated precisely which quantity is uniform and how the abscissa is defined. The figure also lacks axis labels and units for the integral and PDF curves.
[Section IV-A] The phrase 'Since the CDF of L depends on the channel signal-to-noise ratio, can be obtained from (1)' contains a grammatical omission; it should read 'the CDF of L can be obtained from (1)'. This is a minor readability issue.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the fine-tuning method is validated against exact reverse-pair counts and simulated BLER, not against its own fitted parameters.

full rationale

The paper's central claim is a decoded-BLER comparison in Figure 5, obtained by simulating Algorithm 1 on BCH(127,113) and comparing with SGRAND and ORB-type GRAND baselines; no BLER outcomes are used to calibrate the method. The only fitted quantities are the [a,b,c,d] parameters in the approximate position-count function ~O'(m), and these are fitted to the exact O(m) computed from (13), i.e., to the offline basis error-pattern list, not to the reported performance. The reverse-pair estimate I from (16) is checked against the exact expression (12) in Figure 3, with relative error below 1%, and the simplified selection rule (19) is justified by the empirical insensitivity shown in Figure 4(a), then evaluated by external BLER simulation. Theorem 2's use of p_v(O(...))O(...) is an approximation without a stated error bound, and the authors themselves note that the derivation is 'largely heuristic,' but this is a rigor gap rather than a circular reduction: the adjusted positions are not defined so as to force the reported BLER. The self-citation [12] appears only as a comparative baseline (RS-ORBGRAND) in Figure 5 and Table II; no load-bearing premise is justified solely by that citation, and no uniqueness theorem is imported. Every load-bearing quantity (O(m), I, BLER) is either independently computed or externally simulated, so no step reduces by construction to its own input.

Assumptions & free parameters 1 free parameters · 5 assumptions · 0 invented entities

The claim rests on standard partition asymptotics, the AWGN folded-normal model, and several admitted heuristic approximations. The two ad hoc steps are the insensitivity assumption in (19) and the position formula in (18). There are no new physical entities.

free parameters (1)

a, b, c, d (CDF-ORBGRAND O(m) fit) = not reported
Fit \tilde O'(m) = a exp(b m^c)/m^d to the exact position-count O(m) of the CDF-ORBGRAND basis list in Section III-C. These constants affect the reordering in (18), so the algorithm's behavior depends on them; they are not fit to the reported BLER.

assumptions (5)

standard math Szekeres's asymptotic q(n) ~ Q(n) for distinct integer partitions, and replacement of sums by integrals in (14)-(15).
Used to approximate O(m), the number of error patterns with weight at most m. Invoked in Section III-C, Lemma 1 and eq (15).
domain assumption The LLR magnitude L follows a folded normal distribution (eq 1) on AWGN.
Defines the channel model and is used for CDF-ORBGRAND gamma values and simulations. Assumes AWGN with antipodal signaling.
ad hoc to paper The fine-tuned position can be estimated by replacing counts over subsets E_v with p_v(O(...))O(...) in (18).
This heuristic step is used to construct Algorithm 1 and is not rigorously justified; the paper admits the derivation is heuristic. Introduced in Section III-D, Theorem 2.
ad hoc to paper The integral in (16) is insensitive to |δ(u,v)|, allowing replacement by (19).
Empirically observed for D=1 at Eb/N0=7 dB (Figure 4a); load-bearing for simplified selection. Section III-D.
domain assumption ORB-type basis lists satisfy criterion (3) with fixed gamma.
The fine-tuning method is scoped to such lists; Section II-B footnote 1 excludes other structures.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Fine-tuning ORBGRAND with Very Few Channel Soft Values." pith.science (2026). https://pith.science/paper/YL63SNVK

@misc{pith2026250708696,
  author       = {Pith},
  title        = {Pith review of: Fine-tuning ORBGRAND with Very Few Channel Soft Values},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/YL63SNVK}},
  note         = {Machine review of arXiv:2507.08696}
}

read the original abstract

Guessing random additive noise decoding (GRAND) is a universal decoding paradigm that decodes by repeatedly testing error patterns until identifying a codeword, where the ordering of tests is generated by the received channel values. On one hand, while testing error patterns in a descending order of posterior probabilities leads to maximum likelihood decoding, its implementation complexity is prohibitive. On the other hand, testing error patterns with a prescribed set of error patterns permuted by the ranking among magnitudes of log-likelihood ratios (i.e., ordered reliability bits, ORB) enables efficient implementation, but results in performance loss for finite-length codes. Aiming at harnessing the strengths of these two approaches, this work proposes a fine-tuning method to improve ORBGRAND, adjusting the ordering of tests with the aid of very few exact channel soft values. This method is based on a metric for assessing the ``well-orderedness'' of error patterns. The metric is studied via the lens of the asymptotic theory of integer partitioning, which provides highly accurate estimation in numerical experiments. The metric then leads to an effective identification of fine-tuning to conduct, at the cost of a negligible increment of complexity. Numerical experiments demonstrate that the proposed fine-tuning method achieves a substantial performance enhancement compared with ORBGRAND.

Figures

Figures reproduced from arXiv: 2507.08696 by the authors.

**Figure 1.** At t = 4, the loop is repeated three times, and the adjusted positions of e(4), e(5), and e(6) are calculated. E. Complexity Analysis In Table I, we compare the computational complexity of ORB-type GRAND, SGRAND and fine-tuning ORB-type GRAND in the decoding stage. Before decoding, all algorithms have a ranking complexity of O(N log N). And with fine-tuning, (19) needs to be solved for obtaining (d ∗ 1 , ..., d∗ D)… view at source ↗

**Figure 2.** shows the numerical accuracy of O˜(m). From γ, O(m) can be obtained from (13), shown in discrete circles. Since the CDF of L depends on the channel signal-to-noise ratio, can be obtained from (1), we plot the O(m) of CDFORBGRAND at Eb/N0 ∈ {4, 5, 6, 7} dB. For ORBGRAND, we approximate O˜(m) via (15); for CDF-ORBGRAND, we use a·e (b·mc )/md to fit O˜′ (m). We observe from [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗

**Figure 3.** The exact and estimated values of I are plotted as crosses (“x”) and circles (“o”), respectively. The triangle represents the relative error, and the right axis shows that the relative error is much less than 1%. probability that ei(t) = 1 in the first T error patterns in E˜. The figure shows that bits with smaller ranking values are flipped more frequently, and hence p0(T)(1 − p0(T)) in (20) is very sensitive to th… view at source ↗

Figures from the paper (2 more)

**Figure 4.** Figure 4: Analyze (16) under D = 1. (a): The integral value is insensitive to |δ(u, v)|, the ordinates on the left and right are the integral results and PDF value, the i is considered to be uniformly distributed; (b): The Pu(T) is sensitive to the order of the bits. Due to the …

**Figure 5.** Figure 5: BLER for BCH(127, 113), maximum number of tests [PITH_FULL_IMAGE:figures/full_fig_p006_5.png]

Discussion (0). Sign in to comment.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

The Linear Reliability Channel
cs.IT 2025-09 conditional novelty 7.0 of 10

For a new discrete soft-decision channel with linearly ordered reliabilities, the paper derives closed-form error and success exponents for random-code ML decoding and proves soft-decision decoding strictly outperform...

Reference graph

Works this paper leans on

13 extracted references · 13 canonical work pages · cited by 1 Pith paper

[1]

Capacity-achieving guessing random additive noise decoding,

K. R. Duffy, J. Li, and M. M ´edard, “Capacity-achieving guessing random additive noise decoding,” IEEE Transactions on Information Theory , vol. 65, no. 7, pp. 4023–4040, 2019

work page 2019
[2]

Efficient decoders for short block length codes in 6G URLLC,

C. Yue, V . Miloslavskaya, M. Shirvanimoghaddam, B. Vucetic, and Y . Li, “Efficient decoders for short block length codes in 6G URLLC,” IEEE Communications Magazine , vol. 61, no. 4, pp. 84–90, 2023

work page 2023
[3]

Channel coding towards 6G: Technical overview and outlook,

M. Rowshan, M. Qiu, Y . Xie, X. Gu, and J. Yuan, “Channel coding towards 6G: Technical overview and outlook,” IEEE Open Journal of the Communications Society , 2024

work page 2024
[4]

Soft maximum likelihood decoding using GRAND,

A. Solomon, K. R. Duffy, and M. M ´edard, “Soft maximum likelihood decoding using GRAND,” in IEEE International Conference on Com- munications (ICC) , 2020, pp. 1–6

work page 2020
[5]

ORBGRAND is almost capacity-achieving,

M. Liu, Y . Wei, Z. Chen, and W. Zhang, “ORBGRAND is almost capacity-achieving,” IEEE Transactions on Information Theory , vol. 69, no. 5, pp. 2830–2840, 2022

work page 2022
[6]

Lin and D

S. Lin and D. J. Costello, Error Control Coding: Fundamentals and Applications. Englewood Cliffs, NJ, USA: Prentice Hall, 2004

work page 2004
[7]

Ordered reliability bits guessing random additive noise decoding,

K. R. Duffy, W. An, and M. M ´edard, “Ordered reliability bits guessing random additive noise decoding,” IEEE Transactions on Signal Process- ing, vol. 70, pp. 4528–4542, 2022

work page 2022
[8]

High-throughput and energy-efficient VLSI architecture for ordered reliability bits GRAND,

S. M. Abbas, T. Tonnellier, F. Ercan, M. Jalaleddine, and W. J. Gross, “High-throughput and energy-efficient VLSI architecture for ordered reliability bits GRAND,” IEEE Transactions on V ery Large Scale Integration (VLSI) Systems , vol. 30, no. 6, pp. 681–693, 2022

work page 2022

Show all 13 references

[9]

Efficient ORBGRAND imple- mentation with parallel noise sequence generation,

C. Ji, X. You, C. Zhang, and C. Studer, “Efficient ORBGRAND imple- mentation with parallel noise sequence generation,” IEEE Transactions on V ery Large Scale Integration (VLSI) Systems , 2024

2024
[10]

High-performance low-complexity error pattern generation for ORBGRAND decoding,

C. Condo, V . Bioglio, and I. Land, “High-performance low-complexity error pattern generation for ORBGRAND decoding,” in IEEE Globecom Workshops (GC Wkshps) , 2021, pp. 1–6

2021
[11]

Pre-configured error pattern ordered statistics decoding for CRC-polar codes,

X. Li, K. Niu, Y . Han, J. Dai, Z. Tan, and Z. Guo, “Pre-configured error pattern ordered statistics decoding for CRC-polar codes,” Entropy, vol. 25, no. 10, p. 1405, 2023

2023
[12]

Approaching maximum likelihood decoding performance via reshuffling ORBGRAND,

L. Wan and W. Zhang, “Approaching maximum likelihood decoding performance via reshuffling ORBGRAND,” in IEEE International Sym- posium on Information Theory (ISIT) , 2024, pp. 31–36

2024
[13]

Asymptotic distribution of the number and size of parts in unequal partitions,

G. Szekeres, “Asymptotic distribution of the number and size of parts in unequal partitions,” Bulletin of the Australian Mathematical Society , vol. 36, no. 1, pp. 89–97, 1987

1987

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.

[1] [1]

Capacity-achieving guessing random additive noise decoding,

K. R. Duffy, J. Li, and M. M ´edard, “Capacity-achieving guessing random additive noise decoding,” IEEE Transactions on Information Theory , vol. 65, no. 7, pp. 4023–4040, 2019

work page 2019

[2] [2]

Efficient decoders for short block length codes in 6G URLLC,

C. Yue, V . Miloslavskaya, M. Shirvanimoghaddam, B. Vucetic, and Y . Li, “Efficient decoders for short block length codes in 6G URLLC,” IEEE Communications Magazine , vol. 61, no. 4, pp. 84–90, 2023

work page 2023

[3] [3]

Channel coding towards 6G: Technical overview and outlook,

M. Rowshan, M. Qiu, Y . Xie, X. Gu, and J. Yuan, “Channel coding towards 6G: Technical overview and outlook,” IEEE Open Journal of the Communications Society , 2024

work page 2024

[4] [4]

Soft maximum likelihood decoding using GRAND,

A. Solomon, K. R. Duffy, and M. M ´edard, “Soft maximum likelihood decoding using GRAND,” in IEEE International Conference on Com- munications (ICC) , 2020, pp. 1–6

work page 2020

[5] [5]

ORBGRAND is almost capacity-achieving,

M. Liu, Y . Wei, Z. Chen, and W. Zhang, “ORBGRAND is almost capacity-achieving,” IEEE Transactions on Information Theory , vol. 69, no. 5, pp. 2830–2840, 2022

work page 2022

[6] [6]

Lin and D

S. Lin and D. J. Costello, Error Control Coding: Fundamentals and Applications. Englewood Cliffs, NJ, USA: Prentice Hall, 2004

work page 2004

[7] [7]

Ordered reliability bits guessing random additive noise decoding,

K. R. Duffy, W. An, and M. M ´edard, “Ordered reliability bits guessing random additive noise decoding,” IEEE Transactions on Signal Process- ing, vol. 70, pp. 4528–4542, 2022

work page 2022

[8] [8]

High-throughput and energy-efficient VLSI architecture for ordered reliability bits GRAND,

S. M. Abbas, T. Tonnellier, F. Ercan, M. Jalaleddine, and W. J. Gross, “High-throughput and energy-efficient VLSI architecture for ordered reliability bits GRAND,” IEEE Transactions on V ery Large Scale Integration (VLSI) Systems , vol. 30, no. 6, pp. 681–693, 2022

work page 2022