Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Bypass DeepSeek censorship by speaking in hex

substack.com

Post reply on HN
Print view View original

1–10 of 397 posts

Re: Bypass DeepSeek censorship by speaking in hex

#3

Post by tossaway2000 »

> I wagered it was extremely unlikely they had trained censorship into the LLM model itself.

I wonder why that would be unlikely? Seems better to me to apply censorship at the training phase. Then the model can be truly naive about the topic, and there's no way to circumvent the censor layer with clever tricks at inference time.

Re: Bypass DeepSeek censorship by speaking in hex

#4

Post by lxe »

You can also intercept the xhr response which would still stop generation, but the UI won't update, revelaing the thoughts that lead to the content filter:

    const filter = t => t?.split('\n').filter(l => !l.includes('content_filter')).join('\n');

    ['response', 'responseText'].forEach(prop => {
      const orig = Object.getOwnPropertyDescriptor(XMLHttpRequest.prototype, prop);
      Object.defineProperty(XMLHttpRequest.prototype, prop, {
        get: function() { return filter(orig.get.call(this)); }
      });
    });
Paste the above in the browser console ^

Re: Bypass DeepSeek censorship by speaking in hex

#6

Post by ks2048 »

Part of the blog is hypothesizing that the censorship is in a separate filtering stage rather than the model itself. But, the example of hex encoding doesn't prove or disprove that at all, does it? Can't you just check on a version running open-source weights?

Re: Bypass DeepSeek censorship by speaking in hex

#7

Post by noman-land »

lxe wrote: post #4

You can also intercept the xhr response which would still stop generation, but the UI won't update, revelaing the thoughts that lead to the content filter: const filter = t => t?.split('\n').filter(l => !l.includes('content_filter')).join('\n'); ['response', 'responseText'].forEach(prop => { const orig = Object.getOwnPropertyDescriptor(XMLHttpRequest.prototype, prop); Object.defineProperty(XMLHttpRequest.prototype, p…

This is why javascript is so fun.

Re: Bypass DeepSeek censorship by speaking in hex

#8

Post by noman-land »

tossaway2000 wrote: post #3

> I wagered it was extremely unlikely they had trained censorship into the LLM model itself. I wonder why that would be unlikely? Seems better to me to apply censorship at the training phase. Then the model can be truly naive about the topic, and there's no way to circumvent the censor layer with clever tricks at inference time.

I agree. Wouldn't the ideal censorship be to erase from the training data any mention of themes, topics, or opinions you don't like?

Re: Bypass DeepSeek censorship by speaking in hex

#9

Post by foota »

tossaway2000 wrote: post #3

> I wagered it was extremely unlikely they had trained censorship into the LLM model itself. I wonder why that would be unlikely? Seems better to me to apply censorship at the training phase. Then the model can be truly naive about the topic, and there's no way to circumvent the censor layer with clever tricks at inference time.

Probably time to market I would guess?
Post reply on HN