Close Menu
    DevStackTipsDevStackTips
    • Home
    • News & Updates
      1. Tech & Work
      2. View All

      BrowserStack launches Figma plugin for detecting accessibility issues in design phase

      July 22, 2025

      Parasoft brings agentic AI to service virtualization in latest release

      July 22, 2025

      Node.js vs. Python for Backend: 7 Reasons C-Level Leaders Choose Node.js Talent

      July 21, 2025

      Handling JavaScript Event Listeners With Parameters

      July 21, 2025

      I finally gave NotebookLM my full attention – and it really is a total game changer

      July 22, 2025

      Google Chrome for iOS now lets you switch between personal and work accounts

      July 22, 2025

      How the Trump administration changed AI: A timeline

      July 22, 2025

      Download your photos before AT&T shuts down its cloud storage service permanently

      July 22, 2025
    • Development
      1. Algorithms & Data Structures
      2. Artificial Intelligence
      3. Back-End Development
      4. Databases
      5. Front-End Development
      6. Libraries & Frameworks
      7. Machine Learning
      8. Security
      9. Software Engineering
      10. Tools & IDEs
      11. Web Design
      12. Web Development
      13. Web Security
      14. Programming Languages
        • PHP
        • JavaScript
      Featured

      Laravel Live Denmark

      July 22, 2025
      Recent

      Laravel Live Denmark

      July 22, 2025

      The July 2025 Laravel Worldwide Meetup is Today

      July 22, 2025

      Livewire Security Vulnerability

      July 22, 2025
    • Operating Systems
      1. Windows
      2. Linux
      3. macOS
      Featured

      Galaxy Z Fold 7 review: Six years later — Samsung finally cracks the foldable code

      July 22, 2025
      Recent

      Galaxy Z Fold 7 review: Six years later — Samsung finally cracks the foldable code

      July 22, 2025

      Halo and Half-Life combine in wild new mod, bringing two of my favorite games together in one — here’s how to play, and how it works

      July 22, 2025

      Surprise! The iconic Roblox ‘oof’ sound is back — the beloved meme makes “a comeback so good it hurts” after three years of licensing issues

      July 22, 2025
    • Learning Resources
      • Books
      • Cheatsheets
      • Tutorials & Guides
    Home»Development»Machine Learning»LLMs Can Now Retain High Accuracy at 2-Bit Precision: Researchers from UNC Chapel Hill Introduce TACQ, a Task-Aware Quantization Approach that Preserves Critical Weight Circuits for Compression Without Performance Loss

    LLMs Can Now Retain High Accuracy at 2-Bit Precision: Researchers from UNC Chapel Hill Introduce TACQ, a Task-Aware Quantization Approach that Preserves Critical Weight Circuits for Compression Without Performance Loss

    April 22, 2025

    LLMs show impressive capabilities across numerous applications, yet they face challenges due to computational demands and memory requirements. This challenge is acute in scenarios requiring local deployment for privacy concerns, such as processing sensitive patient records, or compute-constrained environments like real-time customer service systems and edge devices. Post-training quantization (PTQ) is a promising solution that allows efficient compression of pre-trained models, reducing memory consumption by 2-4 times. However, current processes have a bottleneck at 4-bit compression, with substantial performance degradation when attempting 2- or 3-bit precision. Most PTQ methods rely on small mini-batches of general-purpose pre-training data to account for activation changes resulting from quantization.

    Current methods for LLM compression primarily fall into three categories. Uniform quantization represents the most basic approach, where weights stored as 16-bit float tensors are compressed by treating each row independently, mapping floats to integers based on maximum and minimum values within each channel. GPTQ-based quantization techniques advance this concept by focusing on layerwise reconstruction, aiming to minimize reconstruction loss after quantization. Further, Mixed-precision quantization methods offer a more nuanced strategy, moving beyond fixed precision for all weights. These techniques assign bit-width based on weight importance to maintain performance, with some approaches preserving high-sensitivity “outlier” weights at higher precision.

    Researchers from UNC Chapel Hill have proposed a novel mixed-precision post-training quantization approach called TaskCircuit Quantization (TACQ). The method shows similarities to automated circuit discovery by directly conditioning the quantization process on specific weight circuits, defined as sets of weights associated with downstream task performance. TACQ compares unquantized model weights with uniformly quantized ones to estimate expected weight changes from quantization, then uses gradient information to predict impacts on task performance, enabling preservation of task-specific weights. TACQ consistently outperforms baselines with the same calibration data and lower weight budgets, and achieves significant improvements in the challenging 2-bit and 3-bit regimes.

    TACQ is defined by a saliency metric that identifies critical weights to preserve during quantization, building on concepts from model interpretability like automatic circuit discovery, knowledge localization, and input attribution. This metric uses two components:

    • Quantization-aware Localization (QAL): Trace how model performance is affected by estimating expected weight changes due to quantization.
    • Magnitude-sharpened Gradient (MSG): A generalized metric for absolute weight importance adapted from input attribution techniques.

    MSG helps stabilize TACQ and addresses biases from QAL’s estimations. These factors combine into a unified saliency metric that can be efficiently evaluated for every weight in a single backward pass, allowing preservation of the top p% highest-scoring weights at 16-bit precision.

    In the challenging 2-bit setting, TACQ outperforms SliM-LLM with absolute margin improvements of 16.0% (from 20.1% to 36.1%) on GSM8k, 14.1% (from 34.8% to 49.2%) on MMLU, and 21.9% (from 0% to 21.9%) on Spider. Other baseline methods like GPTQ, SqueezeLLM, and SPQR deteriorate to near-random performance at this compression level. At 3-bit precision, TACQ preserves approximately 91%, 96%, and 89% of the unquantized accuracy on GSM8k, MMLU, and Spider, respectively, while outperforming the strongest baseline, SliM-LLM, by 1-2% across most datasets. TACQ’s advantages become evident in generation tasks requiring sequential token outputs, where it is the only method capable of recovering non-negligible performance in the 2-bit setting for the Spider text-to-SQL task.

    In conclusion, researchers introduced TACQ, a significant advancement in task-aware post-training quantization. It improves model performance at ultra-low bit-widths (2- to 3-bits) where previous methods degrade to near-random outputs. TACQ aligns with automatic circuit discovery research by selectively preserving only a small fraction of salient weights at 16-bit precision, indicating that sparse weight “circuits” disproportionately influence specific tasks. Moreover, experiments on Spider show that TACQ better preserves model generation capabilities, making it suitable for program-prediction tasks. This also applies to situations involving agents, where models frequently generate many executable outputs, and where efficiency is a concern.


    Check out the Paper and GitHub Page. Also, don’t forget to follow us on Twitter and join our Telegram Channel and LinkedIn Group. Don’t Forget to join our 90k+ ML SubReddit.

    🔥 [Register Now] miniCON Virtual Conference on AGENTIC AI: FREE REGISTRATION + Certificate of Attendance + 4 Hour Short Event (May 21, 9 am- 1 pm PST) + Hands on Workshop

    The post LLMs Can Now Retain High Accuracy at 2-Bit Precision: Researchers from UNC Chapel Hill Introduce TACQ, a Task-Aware Quantization Approach that Preserves Critical Weight Circuits for Compression Without Performance Loss appeared first on MarkTechPost.

    Source: Read More 

    Facebook Twitter Reddit Email Copy Link
    Previous ArticlePonyProg is a serial device programmer
    Next Article Long-Context Multimodal Understanding No Longer Requires Massive Models: NVIDIA AI Introduces Eagle 2.5, a Generalist Vision-Language Model that Matches GPT-4o on Video Tasks Using Just 8B Parameters

    Related Posts

    Machine Learning

    How to Evaluate Jailbreak Methods: A Case Study with the StrongREJECT Benchmark

    July 22, 2025
    Machine Learning

    Building a Smart Python-to-R Code Converter with Gemini AI-Powered Validation and Feedback

    July 22, 2025
    Leave A Reply Cancel Reply

    For security, use of Google's reCAPTCHA service is required which is subject to the Google Privacy Policy and Terms of Use.

    Continue Reading

    Malicious PyPI, npm, and Ruby Packages Exposed in Ongoing Open-Source Supply Chain Attacks

    Development

    CVE-2025-6057 – WordPress WPBookit Arbitrary File Upload Vulnerability

    Common Vulnerabilities and Exposures (CVEs)

    Are software professionals truly an endangered species? It’s complicated

    News & Updates

    Is HR ready for AI?

    News & Updates

    Highlights

    CVE-2025-20965 – Samsung Bixby Unauthenticated Data Access Vulnerability

    May 7, 2025

    CVE ID : CVE-2025-20965

    Published : May 7, 2025, 9:15 a.m. | 2 hours, 20 minutes ago

    Description : Improper handling of insufficient permission in Bixby wakeup prior to version 2.3.74.8 allows local attackers to access sensitive data.

    Severity: 6.2 | MEDIUM

    Visit the link for more details, such as CVSS details, affected products, timeline, and more…

    The Creative World of Srinidhi Ranganathan and Flying Fishes

    April 17, 2025

    How to Switch Screens on Windows: The Complete Guide to Multi-Monitor Productivity 

    June 18, 2025

    How AI Agents Remember Things: The Role of Vector Stores in LLM Memory

    July 17, 2025
    © DevStackTips 2025. All rights reserved.
    • Contact
    • Privacy Policy

    Type above and press Enter to search. Press Esc to cancel.