arXiv subjects

Lixing Li

Publications and source records attributed to Lixing Li.

3 recordsLinked to original sources

arXiv2026

ReconSpan: Reconstruction-Guided Adaptive Latent Tokenization

Adaptive latent tokenization maps a fine-grained input to a shorter sequence of continuous representations associated with input-dependent spans. We introduce ReconSpan, which divides text into chunks that a backward decoder can reconstruct from a single contextual prefix code and retains one such code as the latent token for each chunk. The reconstruction criterion is applied when chunks are formed, allowing one trained autoencoder to produce average chunk lengths from 6.5 to 12.2. At matched average length, reconstruction-guided boundaries preserve more text than random boundaries. Readers of the resulting latent sequence recover topic information reliably but struggle to extract exact details.

cs.CL↗

arXiv2026

Evaluating the Architectural Reasoning Capabilities of LLM Provers via the Obfuscated Natural Number Game

While Large Language Models have achieved notable success on formal mathematics benchmarks such as MiniF2F, it remains unclear whether these results stem from genuine logical reasoning or semantic pattern matching against pre-training data. This paper identifies Architectural Reasoning: the ability to synthesize formal proofs using exclusively local axioms and definitions within an alien math domain, as the necessary ability for future automated theorem discovery AI. We use the Obfuscated Natural Number Game, a benchmark to evaluate Architectural Reasoning. By renaming identifiers in the Natural Number Game in Lean 4, we created a zero-knowledge, closed environment. We evaluate state-of-the-art models, finding a universal latency tax where obfuscation increases inference time. The results also reveal a divergence in robustness: while general models (Claude-Sonnet-4.5, GPT-4o) suffer performance degradation, reasoning models (DeepSeek-R1, GPT-5, DeepSeek-Prover-V2) maintain the same accuracy despite the absence of semantic cues. These findings provide a quantitative metric for assessing the true capacity for mathematical reasoning.

cs.LG↗

arXiv2010

Process Algebra as Abstract Data Types

In this paper we introduced an algebraic semantics for process algebra in form of abstract data types. For that purpose, we developed a particular type of algebra, the seed algebra, which describes exactly the behavior of a process within a labeled transition system. We have shown the possibility of characterizing the bisimulation of two processes with the isomorphism of their corresponding seed algebras. We pointed out that the traditional concept of isomorphism of algebra does not apply here, because there is even no one-one correspondence between the elements of two seed algebras. The lack of this one-one correspondence comes from the non-deterministic choice of transitions of a process. We introduce a technique of hidden operations to mask unwanted details of elements of a seed algebra, which only reflect non-determinism or other implicit control mechanism of process transition. Elements of a seed algebra are considered as indistinguishable if they show the same behavior after these unwanted details are masked. Each class of indistinguishable elements is called a non-hidden closure. We proved that bisimulation of two processes is equivalent to isomorphism of non-hidden closures of two seed algebras representing these two processes. We call this kind of isomorphism a deep isomorphism. We get different models of seed algebra by specifying different axiom systems for the same signature. Each model corresponds to a different kind of bisimulation. By proving the relations between these models we also established relations between 10 different bisimulations, which form a acyclic directed graph.

cs.PL↗