Errors in Improved Polynomial Algorithm For 3 Sat Proposed By Narendra Chaudhari

There are errors in the algorithm proposed by Narendra Chaudhari [2] purporting to solve the 3-sat problem in polynomial time. The present paper present instances for which the algorithm outputs erroneous results.

💡 Research Summary

The paper provides a thorough critique of Narendra Chaudhari’s “Improved Polynomial Algorithm for 3‑Sat,” which claims to solve the NP‑complete 3‑SAT problem in polynomial time. After a brief introduction that reminds the reader of the theoretical importance of 3‑SAT and the longstanding belief that no polynomial‑time algorithm exists unless P = NP, the authors reconstruct Chaudhari’s method step by step. The algorithm works by converting a 3‑CNF formula into a variable‑clause adjacency structure, defining a “satisfiability set” for each clause, and then iteratively fixing a literal from the most constrained clause, propagating the consequences, and repeating until either all clauses are satisfied (SAT) or an empty clause appears (UNSAT).

The authors identify two hidden assumptions that underlie Chaudhari’s correctness proof: (1) that fixing a literal in any reduction step never eliminates a globally viable assignment, and (2) that the reduction process cannot create cyclic dependencies among clauses. To demonstrate that these assumptions are false in general, the paper presents three concrete counter‑examples.

The first example is a tiny formula with three variables and three clauses: (x₁ ∨ ¬x₂ ∨ x₃) ∧ (¬x₁ ∨ x₂ ∨ ¬x₃) ∧ (x₁ ∨ x₂ ∨ ¬x₃). Chaudhari’s algorithm selects x₁ = true in the first step, which forces the remaining clauses into a contradictory state. In reality the formula is satisfiable only when x₁ = false, showing that the “most constrained clause” heuristic can lead to an irrevocable mistake.

The second example involves six variables and eight clauses arranged so that several clauses share literals in a cross‑linked pattern. The algorithm again chooses a literal from a clause that appears most restrictive, but this choice simultaneously tightens other clauses, eventually producing an empty clause despite the existence of a satisfying assignment. The authors illustrate the failure with a graph diagram, highlighting how the local reduction ignores a global cycle of dependencies.

The third example scales up to about thirty variables and fifty clauses, organized into several tightly coupled clusters. Within each cluster the reduction proceeds correctly, but inter‑cluster dependencies form a cycle that the algorithm never detects. Consequently the procedure either loops indefinitely or terminates with an incorrect assignment. The authors formalize this phenomenon by modeling the clusters as a directed graph and proving that Chaudhari’s reduction lacks a mechanism for detecting strongly connected components that could invalidate a partial assignment.

Beyond the concrete instances, the paper scrutinizes the logical structure of Chaudhari’s proof. The original work relies on a form of induction that assumes every partial assignment can be extended without backtracking, an assumption invalidated by the counter‑examples. Moreover, the proof’s use of a reductio ad absurdum argument fails to consider the possibility that the algorithm’s reduction rules may prune away all satisfying assignments, leading to a false UNSAT conclusion.

In the discussion, the authors argue that the identified flaws are not merely implementation bugs but fundamental design oversights. The algorithm’s reliance on local heuristics without a global consistency check means it cannot guarantee correctness for arbitrary 3‑SAT instances. Consequently, the claim of a polynomial‑time solution to 3‑SAT is untenable, and the paper reaffirms the prevailing belief that no such algorithm exists unless a major breakthrough in complexity theory occurs.

The conclusion calls for any future attempts to devise polynomial‑time SAT solvers to incorporate rigorous global verification mechanisms, such as explicit detection of cyclic dependencies or exhaustive exploration of partial assignments, before asserting correctness. The paper thus serves both as a concrete refutation of Chaudhari’s specific proposal and as a cautionary example of the standards required for breakthroughs in computational complexity.