CHASD: Language Increment-Calibrated Contrastive Decoding against Hallucination in LVLMs

Dimensions overall score 4.0

• An inference-time framework for Large Vision-Language Models that reduces object hallucinations by selectively applying contrastive decoding based on token-specific confidence. Training-free contrastive decoding methods mitigate this issue by comparing predictions from original and perturbed visual inputs, but existing approaches either apply global perturbations that may alter useful visual evidence or invoke an additional negative branch at every decoding step.problem 35%
  Evidencepartial

  An inference-time framework for Large Vision-Language Models that reduces object hallucinations by selectively applying contrastive decoding based on token-specific confidence. Training-free contrastive decoding methods mitigate this issue by comparing predictions from original and perturbed visual inputs, but existing approaches either apply global perturbations that may alter useful visual evidence or invoke an additional negative branch at every decoding step.

  Implicationpartial

  Abstract-backed fallback claim; anchored extraction has not materialized a public claim row yet.

  Verificationpartial

  partial

• Large Vision-Language Models have shown strong multimodal reasoning capabilities, yet they remain susceptible to object hallucinations when language priors dominate insufficient or misaligned visual evidence. Training-free contrastive decoding methods mitigate this issue by comparing predictions from original and perturbed visual inputs, but existing approaches either apply global perturbations that may alter useful visual evidence or invoke an additional negative branch at every decoding step.method 35%
  Evidencepartial

  Large Vision-Language Models have shown strong multimodal reasoning capabilities, yet they remain susceptible to object hallucinations when language priors dominate insufficient or misaligned visual evidence. Training-free contrastive decoding methods mitigate this issue by comparing predictions from original and perturbed visual inputs, but existing approaches either apply global perturbations that may alter useful visual evidence or invoke an additional negative branch at every decoding step.

  Implicationpartial

  Abstract-backed fallback claim; anchored extraction has not materialized a public claim row yet.

  Verificationpartial

  partial

• ScienceToStartup currently rates this 4.0/10 on the public viability pass. Experiments on POPE, AMBER, MME, MMHal-Bench, and CHAIR show that CHASD improves hallucination-related metrics over strong training-free baselines with competitive inference efficiency.result 35%
  Evidencepartial

  ScienceToStartup currently rates this 4.0/10 on the public viability pass. Experiments on POPE, AMBER, MME, MMHal-Bench, and CHAIR show that CHASD improves hallucination-related metrics over strong training-free baselines with competitive inference efficiency.

  Implicationpartial

  Abstract-backed fallback claim; anchored extraction has not materialized a public claim row yet.

  Verificationpartial

  partial

• LLM Applications moved forward this cycle; last verified May 2026. Public score 4.0/10.why now 35%
  Evidencepartial

  LLM Applications moved forward this cycle; last verified May 2026. Public score 4.0/10.

  Implicationpartial

  Abstract-backed fallback claim; anchored extraction has not materialized a public claim row yet.

  Verificationpartial

  partial

Competitive landscape

An inference-time framework for Large Vision-Language Models that reduces object hallucinations by selectively applying contrastive decoding based on token-specific confidence.