Skip to main content

Envisioning is an emerging technology research institute and advisory.

LinkedInInstagramGitHub

Since 2010

research
  • Observatory
  • Adaptive capacity
  • Newsletter
  • Methodology
  • Origins
  • Vocab
  • RSS feeds
services
  • Signals Session
  • Bespoke Projects
  • Build Sessions
  • Pricing
  • Use cases
  • Signals
  • Signal Scan↗free
impact
  • ANBIMAFuture of Brazilian Capital Markets
  • IEEECharting the Energy Transition
  • Horizon 2045Future of Human and Planetary Security
  • WKOTechnology Scanning for Austria
solutions
  • Innovation
  • Strategy
  • Consultants
  • Foresight
  • Associations
  • Governments
  • L&D
resources
  • Partners
  • Coding for Non-Coders
  • How we work
  • Data visualization
  • Multi-Model Convergence
  • FAQ
  • Security and privacy
  • Public sector
about
  • Manifesto
  • Community
  • Events
  • Support
  • Contact
ResearchCapabilityServicesSignalsAbout
ResearchCapabilityServicesSignalsAbout
  1. Home
  2. Vocab
  3. Distillation Tax

Distillation Tax

Performance ceiling when training smaller models from larger model outputs

Year: 2024Generality: 519Added: Apr 1, 2026
Back to Vocab

Distillation Tax refers to the performance gap between a student model trained through knowledge distillation and the teacher model from which it learns. Knowledge distillation trains a small model to mimic a large one and is a common approach for creating deployable models. Empirical evidence shows that student models consistently underperform their teachers, and the gap can persist regardless of training technique or data quantity. This gap is the distillation tax. It is a structural limitation: the student inherits the teacher's knowledge along with its blind spots, failure modes, and implicit biases.

The student learns a compressed approximation of the teacher's behavior. The teacher encodes capabilities and knowledge in its weights; distillation attempts to transfer this through soft targets (probability distributions) and matching hidden states. The student has fewer parameters, lower capacity, and must make tradeoffs about what to memorize versus what to generalize. When the teacher makes a systematic error or has a particular weakness, the student often amplifies it. The student also cannot invent knowledge the teacher lacks; it is upper-bounded by the teacher's competence. Even with perfect knowledge transfer, the student's smaller capacity prevents it from expressing everything the teacher knows. A 7B model trained on outputs from a 70B teacher will have a performance ceiling below the teacher, no matter how careful the distillation process.

The tax matters for AI development strategy. As models scale up, there is pressure to create smaller, cheaper, faster versions. Distillation is a key technique for this, yet the tax means efficiency comes at a cost. Organizations must decide whether a distilled model is sufficient for their use case or whether they need the full capability. The existence of a fundamental tax also points to a limitation of pure distillation; capable, efficient models may require alternatives such as architecture innovation, different training paradigms, or accepting the tradeoff as the cost of deployment.

Research this in Signals

Scan Distillation Tax for yourself.

Signals turns a topic into a sourced research record you can inspect and rerun. Your first scan is free, and this one starts with Distillation Tax already loaded, so edit it or scan as is.

Related

Related

Distillation
Distillation

Compressing a large teacher model's knowledge into a smaller, efficient student model.

2015Generality: 792
Model Distillation
Model Distillation

A compression technique that trains a small student model to mimic a larger teacher model.

2015Generality: 713
Teacher Model
Teacher Model

A large, pre-trained model that transfers knowledge to a smaller student model.

2015Generality: 620
Alignment Tax
Alignment Tax

Performance cost of making AI models safer and aligned with human values

2022Generality: 693Added: Apr 1, 2026
TAID (Temporally Adaptive Interpolated Distillation)
TAID (Temporally Adaptive Interpolated Distillation)

A distillation technique that aligns teacher and student models across differing temporal resolutions.

2022Generality: 380
Saturation Effect
Saturation Effect

Diminishing performance returns as model complexity or training data increases beyond a threshold.

1986Generality: 590